PSRC Data Stories

Shiny Data Visualization for Urban Planning

Thu, 31 Oct 2024 00:00:00 -0700

Why do we use Shiny to visualize urban planning data?

Over the past several years, the PSRC data science team has deployed several Shiny apps that create interactive web data visualizations to inform regional urban planning.

The shiny apps are created to eludicate the complex interactions between demography, economy, land use and transportation.

We recommend other urban planning data experts who program in python and R like us check out Shiny as option. Our team uses many other data visualization software packages such as Tableau, Infogram, ArcGIS Story Maps, and yes, Excel. Shiny stands out because it allows for an open source clean integration of code along a data pipeline from engineering, to statistics, and into a final product, without a ton of web programming expertise. All of our code and data to create the apps is open!! See the our shiny app code here on Github.

We also developed an R library to easily create nicely formatted charts so we can quickly go from numbers to meaningful data visualization without mucking around with getting the colors, fonts, and other layout parameters set by hand. Here you can see some of the color scales from the library.

We host our Shiny apps on shiny.apps.io so we don’t have worry about server details. Shiny.apps.io has free hosting service, but we pay about $500 a year to get more features like unlimited apps and performance. We’ve got about 20 apps up there!

We’ve been developing our coding and Shiny chops on the urban planning data team for about the past ten years, so using Shiny wasn’t an overnight solution. As an example of the continous skill-building our team does, this year several team members attended the Posit Conference to increase our Shiny and other programming skills in python and R. (Posit is the company that develops Shiny.)

Shiny apps work well for our situation that we have statisticians, GIS experts, and data analysts with a little coding skill who want to push our data into an open venue. If you’re a real web developer with skills in javascript and css, Shiny also might not be the thing for you.

Top Seven PSRC Shiny apps

Click on the links to see the apps live.

Regional Transportation Plan Dashboard
The regional transportation plan dashboard is my favorite Shiny app because it integrates with a top planning process at PSRC. Historically, the regional transportation plan has been a hundreds of pages long PDF. The Shiny app brings the transportation plan to life with relevant, easily understandable data in a visually appealing framework. We toiled away to make the graphics match our website and have a modern design.

Seattle Open Data Hackathon Housing Permit App

Christy Lam, Brandon Wong, and Joanne Lin participated in Seattle's Open Data Hackathon.

Last year several data team members participated in the City of Seattle's open data hackathon. The team developed an app that shows time to complete housing permitting processes in Seattle's Racial Social Equity areas. We got second place! I like this app because it shows how teamwork makes the data science dream work.

Household Travel Survey Explorer

Household travel survey results showing that workers who work at home drive fewer miles. The app will show this data eventually. For now, I just want this data out because it's important to policies being enacted today!

The household travel survey explorer creates data summarizations from our rich dataset of household's daily activities. The app summaries show the diversity of travel behavior by different demographic and geographic characteristics. Learn more about the survey program.

Community Profiles
The community profiles app allows local jurisdictional planners to quickly access key Census and American Community Survey data from their area. Census data underlies much of the work we do as planners, but it can be inaccessible at times, and this app puts the important information at planner's fingertips.

Summer Planning Academy Toolbox

High school students used a shiny app we developed to learn about urban planning. You can see images they took from the app in the presentation behind the students.

The summer planning academy at PSRC introduced high school students underrepresented in urban planning to our work. The Shiny app we developed packaged together maps and data to help the students create a capstone project. We were wowed to see what they were able to do with easy access to the data provided in the app. They learned a ton about planning data!

Centers Monitoring Dashboard
PSRC's regional centers are areas of the region targeted to add new housing and job density. I like this dashboard because regional centers are a key aspect of PSRC's land use planning paradigm, and the dashboard allows us to track center performance.

CTPP Hackathon App

A screenshot of the app we developed to show Census Transportation Planning Package data in regional centers.

The Census Transportation Planning Package(CTPP) is a dataset that allows for a deep understanding of transportation data coming out of the Census. The CTPP team was developing an API, and we helped them test it by developing an app. We won first place in a national competition! The main reason I like this app is: teamwork makes the dream work again, and we helped out the CTPP national development team by testing (we did a public service!)

Hoping to inspire others!

The PSRC data team has reached to integrate programming, statistics, and urban planning to create more modern data products for about ten years now.

We’ve learned a ton along the way and we’d love to share. Here’s some hard-earned wisdom:

Fully preprocessing data to feed to the app saves valuable user time as opposed to performing analytical calculations in the app
90% of app development is graphical formatting -getting things just right. When you get it just right, the response is exponentially higher than just conveying the raw information without nice visuals.
Bringing an almost fully baked app at least visually (if not functionally) to people before we get feedback is critical- otherwise people don’t understand our vision.

Do you want to learn more? Or do you have something you’ve learned you’d like to share? Please reach out! email me: Suzanne Childress

Shiny For Planning

Thu, 31 Oct 2024 00:00:00 -0700

Shiny Data Visualization for Urban Planning

Shiny for Urban Planning Data

Over the past several years, the PSRC data science team has deployed several Shiny apps to inform regional urban planning.

For the past x years building expertise- software carpentry

hackathon picture

mutiple awards for our shiny apps

list the apps

Posit conference host on shiny.apps.io so we don’t have to maintain a server

what we’ve learned

we also use story maps, infogram, dash;

python stack

https://psrcwa.shinyapps.io/rtp-dashboard/ https://psrcwa.shinyapps.io/community-profiles/

https://psrcwa.shinyapps.io/centers-monitoring/

Survey Data Analytics https://psrcwa.shinyapps.io/travel-survey-explorer/

Summer Planning Academy https://psrcwa.shinyapps.io/planning-academy-centers-2024/

Hackathons https://psrcwa.shinyapps.io/seattle-hackathon/ https://psrcwa.shinyapps.io/ctpp-explorer/

Conclusions

The Puget Sound Regional Household Travel Survey contains detailed information about travel behaviors by race, ethnicity, gender, income, and location. The rich travel survey data source reveals the diversity women’s experiences of travel throughout the region. Through the analysis of the Regional Household Travel Survey in this report, key trends have emerged that differentiate women resident’s travel patterns from men’s travel patterns across all modes. The 2017, 2019, and 2021 household travel surveys collected day-to-day information from households in the central Puget Sound region, such as how we traveled, where we went, how long it took—even where we chose to live and whether we got home deliveries, prior to COVID-19 and after.

Women's transportation needs

Tue, 28 Mar 2023 00:00:00 -0700

Women’s Diverse Travel Needs Often Go Overlooked

Women have transportation needs that have not historically been met in urban environments. The regional transportation system was traditionally designed around the need to quickly get to work at central locations at peak periods. Women tend to carry significantly more of the care-giving burdens of society and thus need a system that works safely for traveling with others to a variety of locations at all times of day. Women also represent a large portion of older populations who have unique transportation needs that are not well-served by our system built around driving and work destinations. Women, especially women of color, are more likely to live in poverty and have lower earnings for the same jobs as men, which impacts many aspects of transportation needs.

With the arrival of the COVID-19 pandemic, we saw a greater trend towards telecommuting for some people and continued trends in aging. Housing prices continue to soar and childcare options continue to degrade. With these new considerations at play, compounded with the existing burdens on women, when will the transportation system evolve for women across diverse backgrounds?

The Puget Sound Regional Council Household Travel Survey provides data on travel by women

This report starts with comparing travel by gender, during the stable period prior to COVID-19, and then dives into some broader, national trends, and finishes by taking a glance at recent trends that occurred in 2021, during COVID-19. Data from 2017 and 2019 have combined to give a more robust sample size. You can also view the full dataset here including 2017, 2019 and 2021 data.

Travel demands put different pressures on women

Women’s trips are more varied to a broader spread of destinations and are more likely to primarily serve the needs of someone else. Women are more likely to live in a car-free or car-light household, take more trips with other people, take fewer single-occupant car trips than men, and are more likely to carpool or get a ride from a family member or friend if they don’t have a driver’s license.

common mode of transportation for children under 3: carried by parents

Women in households with more than two people tend to travel more with other people than men. Transit is often not well set up for people who are traveling with strollers. The walk and bike network are not built out for people of all ages to use.

Gender and race intersect in transportation needs

Some women of color experience poverty and low auto ownership at much higher rates than others. Women in some racial groups experience greater disparity in earnings compared to other racial groups. Other women experience greater differences compared to men. Learn more about how we can redesign and shape our cities through a more equitable process in Leslie Kern’s book entitled Feminist City.

Women of color use transit most

Women of color in the region used transit on more of their trips than men and white, non-Hispanic women-prior to COVID-19. Women of color used transit on nearly double the share of their trips as compared to white, non-Hispanic women. Women of color most likely used transit more than white, non-Hispanic women because of having lower incomes, fewer autos, and living in more transit-friendly areas, but more data is needed to confirm this hypothesis.

Women bike far less than men

Women bike much less than men, regardless of race and ethnicity or income. Some of this is undoubtedly because of bike network design does not feel well-suited or safe for all people of different ages and abilities. There are many considerations as to why women do not feel comfortable riding bicycles as a form of transit, as seen in a study performed by the Institute for Transportation and Development Policy in 2022. The inclusion of designated bike lanes, not only that increase comfort and safety for individuals, but also make traveling with children or larger cargo bikes easier would help create more accessibility and user-friendly options. Additionally, increased education about bike routes and access to bikes as a means to developing a biker’s comfort and familiarity with bike safety, would help aid in bike commuting ease.

As a result of some of these considerations not being implemented, in 2019 only 6,000 bike trips were made by non-binary folks in the central Puget Sound region, and 30,000 bike trips were made by women, but 74,000 were made by men.

Not only are certain modes not conducive to the ways in which women move about their environment, but there is also the concern for safety. The Institute for Transportation and Development Policy in 2022 reported that even though women chose biking less often than men, there was still an active participation in dialogues about how to increase user uptake and accessibility with biking because they deemed it a way to remain safe from harassment or negative transit or walking encounters that they experienced. Women should not have to choose between safety concerns.

How can the transportation and land use system better accommodate older people, household maintenance and caregiving needs? Addressing these questions will improve the system for all genders.

Regional and nationwide data show similar constraints

Women represent more of the older population who need specialized transportation

The findings from Understanding How Women Travel about women’s mode choices, how likely they are to travel with others in their care, and their complex trip-chaining patterns, could all inform adjustments to Metro’s fare policy to make it more equitable towards women and more cost-competitive with driving and carpooling. Women take shorter more local trips, as opposed to long highway trips. In the central Puget Sound region, women’s median trip distance is around 3 miles, whereas men’s median trip distance is around 4 miles (source: 2017/2019 Household Travel Survey). Findings about women’s trip purposes and primary responsibility for household errands could all inform the way transit vehicles, transit stations, and bus stops are designed, so that space for traveling with others and carrying belongings could be better accommodated. Findings about when women are traveling, and average trip lengths could inform new service offerings that meet a mid-day peak travel demand and more local service.

Another consideration for amending transit and transportation access is that, on average, women tend to live longer than men, so that at older ages there are many more women who have unique travel needs than men. Women represent a greater share of the older population who needs more specialized transportation services, and less service to work locations.

Older Populations by Sex in the Puget Sound Region

Changes brought about by COVID-19 pandemic

Although there were apparent disparities in the 2017/2019 data, COVID-19 brought about abrupt changes to the transportation landscape in 2021. The 2021 Household Travel survey showed big changes in travel behavior. More people were walking and biking, and fewer people were using transit. Many more people also now had the opportunity to telework. Although women teleworked more than men before COVID-19, the data shows this increased even more in 2021. In 2021, 41% of women teleworked, as compared to 33% of men. The difference between telework rates relates to women’s job sectors as compared to men, and also household responsibilities. Women of color were about as likely to telework as white, non-Hispanic women.

Telework rates skyrocketed after COVID-19

Source: PSRC Travel Survey

Women of color continued to use transit more during COVID-19

Post-COVID-19, people of color continued to use transit more than white, non-Hispanic people. Women of color were still more likely than white, non-Hispanic women to ride transit. Most likely the women of color who still used transit lacked other options, but more investigation is needed. The largest drop in transit mode shares were for white, non-Hispanic men.

Women of color use transit more

Source: PSRC Travel Survey

Trip purposes by gender changed from 2017/2019 to 2021

Not only were there changes in the trips made to work for men and women, but overall trip purpose changed from 2017/2019 to 2021 (Figure 9). COVID-19 brought about an increase in social, recreational, or restaurant/meal pick-ups in 2021 for both men and women, but women surpassed men in the percentage of these trips made. For more information or differences seen in deliveries versus trips made, visit PSRC’s website for a more in-depth analysis. Additionally, although it appeared as though women made fewer trips escorting additional family members (from 13% to 8%), men’s trips remained at about 7%, which means that overall trips changed, possibly due to the increase in women’s telecommute behaviors or lack of activities outside of the home. Overall, women made more trips than men in all other categories besides work or school trips.

Trip Purposes shifted post COVID-19

Source: PSRC Travel Survey

Conclusion

What is the future of transportation for women?

With increased care-taking demands, and with women more likely to live in poverty and having less employment and lower incomes than men, there are many aspects of transit and transportation modes that need to be reassessed to support women’s needs. Women of color need a robust transit network with more reliable service throughout the day. Biking, walking, and transit don’t always feel accessible or safe, and access to a vehicle isn’t always available to women. While an increase in telework has removed some of the trip-chaining burden for women, we need to consider how we create and manage transportation so that it may be equitable to all. Transit, especially, needs to be designed around women of color, since they use transit more than other groups. While we are working towards a more non-binary understanding of gendered data, we acknowledge that much of the data presented above does analyze information and patterns in a binary manner.

Download the data used in this Trend.

The Diversity of Asian and Pacific Islander Experiences

Wed, 11 May 2022 00:00:00 -0700

Across Heritages and Where People Live

And the challenges but importance of racial equity data

The Puget Sound Region has about 600,000 people who identify as Asian Alone and more than 35,000 people who identify as Native Hawaiian and Pacific Islander Alone, as we discovered last year during Asian American and Pacific Islander heritage month. Asian, Native Hawaiian, and Pacific Islander groups represent a wide variety of many different heritages in the region.

This year for Asian American and Pacific Islander month we wanted to dive deeper to learn more about the how Asian and Pacific Islander experiences vary across different groups based on their heritage and where they live in our region. This analysis results from a top agency priority of advancing racial equity.

Aggregate Census data can tell one story about people’s experiences, but underneath that lies a wide variety of individual experiences. Broad data about race and ethnicity can at times hide how different people really experience the world.

While writing this post, I learned that Native Hawaiian and Pacific Islander people experience significantly more disadvantages than some Asian groups. For future analyses, I plan to keep the two groupings separate based on our findings.

How do worker earnings for Asian and Native Hawaiian Pacific Islander people compare to other groups?

Asian workers as an overall group have one of the highest median personal earnings at $53,000 annually. On the other hand, Native Hawaiian and Other Pacific Islander alone workers have some of the lowest median earnings at $32,000.

How do worker earnings vary across different groups of Asian, Native Hawaiian, and Pacific Islander people?

Median earnings vary widely across different groups of Asian, Native Hawaiian, and Pacific Islander people. Burmese, Other Native Pacific Islander, Nepalese, Fijian, Marshallese, and Hmong workers in the region have median earnings of less than $30,000 per year. On the other end of earnings, Asian Indian workers have median earnings of over $97,000 per year.

detail: PUMS: RAC2P, PERNP Fields, workers working more than 10 hours per week

Asian, Native Hawaiian and Pacific Islanders education levels

Similarly to median earnings, Asian Alone people as a broad group have some of the highest education levels in the region by race, with 56% of adults over 25 having obtained a bachelor’s degree. Only 11% of Native Hawaiian and Pacific Islander adults over 25 had obtained a bachelor’s degree.

Detail: ACS Table C15002D: Educational Attainment by Race (2015-2019)

How do education levels for Asian and Pacific Islander adults vary across the region?

We were interested in looking into how educational levels vary across the geographies of the region. The map below shows that the share of Asian alone adults over 25 with a Bachelor’s degree or higher ranges vary broadly across the region[1].

Over 60% Asian Alone adults in King County have a bachelor’s degree, as compared to 30% of Asian Alone Adults in Pierce County.

Native Hawaiian and Pacific Islander adults have low levels of education throughout the region, as around 10% of adults have a bachelor’s degree or higher.

detail: ACS Table C15002D: Educational Attainment by Race (2015-2019)

Intersecting diverse heritage and geography

Where do different groups live in our region? How does where different people live relate to differing levels of earnings and education for Asian and Pacific Islander people geographically?

Different groups of Asian people have settled in different parts of the region. Those groups experience different levels of earnings and education. As a result of where different groups have settled, in some places in the region Asian people experience high levels of education and income, but in other places Asian people experience low levels of education and income.

The map below shows the top Asian alone group by Census tract in the region, for those tracts that have more than 200 people in that top group.

Asian alone people as a total group in King County have high earnings and education. (And yet again, within King County there is high variation). The top Asian groups in King County are Chinese, except Taiwanese,and Asian Indian people – the groups with high incomes.

Some of groups with the lowest earnings have big populations in Pierce County, and hence the median income is lower in Pierce County.

detail: Table B02015 ACS 5-year 2015-2019

The geographic distribution of people’s heritages impacts where Asian people have greater advantages and disadvantages. When planning for racial equity, we must consider the diversity of where people live and where they come from, and look beneath the surface of regional medians and average. On the data side, we grapple with the low sample sizes and often fear reporting.

Using Statistical Evidence to Explain Experiences of Smaller Populations

How do we report data on important groups when the sample sizes are small?

When making this analysis, I made several statistical choices in reporting. I could have not chosen to not create the two maps that show the geographic distributions because of low sample sizes by tract.

As data analysts, we can choose to not report the racial equity data at all for fear of making a statistical error. But if we only report data for large groups of people, we lose the important experiences of the few.

So in this case, I chose instead to push the data to a reasonable limit. I still wanted to use statistical evidence not statistical significance to make choices, but allow flexibility so that I could build some level of understanding.

On the working earnings charts, I reported the margins of error to show that overall there is very clearly a wide gap in working earnings across groups. Unfortunately, I have found that with even with an educated audience the error bars often do not communicate this variation. By reporting the results, I’m accepting the fact that the statistical results can be possibly misinterpreted as stronger than they are because I think it is more important to highlight the disparity than to ensure that the results are correctly interpreted by all audiences.

For the map of Bachelor’s degree shares, I made several judgment calls to be able to report the data. Here was my thought process: to determine the threshold for reporting, I found that the ACS has roughly a 10% sample rate of the population. If the population of adults is 300 by tract then a conservative margin of error is around 15%; it looks like the map ranges from shares around 10% to 97% so there is truly a wide range. For data storytelling, I chose to use a population of 300 adults as a threshold for reporting. Areas of the map with less than 300 Asian Alone adults were gray. I’m also 100% sure there is a more robust way of doing this analysis, but I also have to balance my time with getting information out there. I could spend weeks trying to do things right and only come out with one data story a year instead of twenty.

Finally, for the map of where different Asian alone groups live, my top priority was to learn where some disadvantaged groups settled in the region. I truly wanted to understand where different heritages live because I am ignorant. So really it was a judgment call based on what is reasonable to me. The tricky thing is there are so many different groups of people and small sample sizes. My compromise was to map the group with the highest population size, and only if that population is greater than 200 people. This approach is problematic for several reasons. Showing only the highest group does not show the number of Asian people overall. There could be a tract with a high number of ten different groups of people, and the data only shows one of those groups. On the other hand, there could be a tract that is 100% one group, and the data doesn’t show that either. I probably should have added more detail to the map, but I left it as is, for simplicity and to push the data out.

What should we do long term to report data for diverse groups?

Continue to push for larger samples sizes for diverse groups. This can mean orders of magnitude larger budget requirements.
Supplement statistical data with individual level experience data to highlight the variation within groups. I’d love to learn more about how I can do this! It feels challenging at the moment because I do not have the tools to gather people’s experiences, but I do have the tools to gather data sets.

Finally, I enjoyed this article about overcoming the small-n problem and recommend it if you would like to learn more.

Technical notes: For these analyses, we used an R package PSRC data staff built called psrccensus that makes it’s really easy to pull ACS, PUMS, and Census data for our region. The code for this analysis is here: https://github.com/psrc/heritage-month/blob/main/aian_2_history_month_region.Rmd

Also I had to learn a ton about including interactive R data visualizations on github pages with Jekyll for this post. The best information for this came from this post: https://jayrobwilliams.com/posts/2020/09/jekyll-html. Thanks Rob Williams!

[1] Because of data availability at a small geographic level, we had to focus on Asian Alone adults only, and did not have sufficient data to analyze Native Hawaiian and Pacific Islander adults.

Discovering Household Travel Data Stories Through Collaboration

Thu, 11 Mar 2021 00:00:00 -0800

In 2020 when we were all starting to work at home, the PSRC data team formed a working group to learn from each other how Household Travel Survey Data could be used to inform urban planning policy. We have used this COVID-19 period to grow our skills in working with household travel survey data.

Thank you to Abdi Mohamud, Brian Lee, Christy Lam, Drew Hanson, Grant Gibson, Neil Kilgren, Polina Butrina, and Mary Richards who shared their brains and time in this effort!

We use GitHub to share scripts with each other.

Source: Our Household Travel Survey GitHub repository

Upskilling Before Discovering

We made many exciting data discoveries during the past year, but before we could really dive into the data, we needed to increase our statistics and scripting skills specific to working with survey data.

We did statistics, learning how to apply sample weights and calculate margins of error.

We wrote functions to easily summarize the data and collaborated via GitHub. Here’s an example of a function we shared with each other:

# Create margins of error for dataset
categorical_moe <- function(sample_size_group){
  sample_w_MOE<-sample_size_group %>%
    mutate(p_col=p_MOE) %>%
    mutate(MOE_calc1= (p_col*(1-p_col))/sample_size) %>%
    mutate(MOE_Percent=z*sqrt(MOE_calc1))
  
  sample_w_MOE<- select(sample_w_MOE, -c(p_col, MOE_calc1))
  
  return(sample_w_MOE)
}   

Data Discoveries

After we did some upskilling, we were ready to make discoveries on a wide range of hot planning topics like residential displacement, travel by people of color, teleworking, home deliveries, and new mobility.

Discovery Park

Source: a picture I took a few years ago

Residential Displacement

A new question that was asked on the 2019 survey was why people moved from their previous homes. We discovered that:

About a quarter of households that moved in the past five years within the four-county region reported that they relocated for one or more negative reasons (also called displacement): cost of housing, forced to move, change in income, or loss of community. Cost of housing was by far the top reason.
One in five white households (22%) felt pressured to leave their homes, as did 14% of Asian households. But nearly one in three (30%) Other People of Color households moved elsewhere because they had to.
We found that displacement is truly a regional phenomenon. Across the region and its counties, displacement was reported at statistically the same rate (within the margin of error).

If you want more info, we made blog posts about our findings on residential displacement. The Seattle Times wrote story using our data.

Understanding travel by different racial groups

Racial equity is a key part of PSRC’s planning efforts. To analyze people’s travel by race using the survey data we had to figure out how to balance adequate sample sizes with richness of detail about people’s race. At first, we categorized people into three broad groups: Asian Only, non-Hispanic White, and Other People of Color, as in this blog post about travel patterns by race. For example, we reported that “Other People of Color were the most likely to take transit to work (17%), followed by Asians (11%) and whites (9%).”

We realized that although our sample sizes were small, we needed to try to report African American transportation experiences separately because they are so unique and because of the history of oppression on our county. In this later blog post, we specifically analyzed travel by African American people separately. We reported that:

“The survey showed that African American people were much more likely to live in a household with no cars than Non-Hispanic Whites. Around one out of five of African American people lived in a household with no cars, with only about one out of twenty Non-Hispanic Whites having lived in a household with no cars. Asian, Hispanic, and Other race people’s travel behavior generally fell between that of Non-Hispanic Whites and African American people. For example, African American people used transit on 12% of their trips, Asian people used transit on 8% of their trips, Hispanic and Other people use transit on 5% of their trips, and Non-Hispanic Whites use transit on 4% of their trips.”

University of Washington has a diverse student population. The region will continue to become more diverse over time. We need to represent this diversity in our data.

Source: PSRC Flickr account:https://www.flickr.com/photos/soundregion

Also, we encountered a puzzle in terms of how to describe household races: How do we describe a household’s race? Not only are people multi-racial, households are also multi-racial. Residential choices and vehicle ownership choices are really made at a household level, as opposed to by individuals in the household, and so we need to summarize data at a household level. The Census uses this antiquated term of householder to describe household races.

Our work assigned households to racial categories as follows. Based on the composition of adults and the person-level categorization, each household was assigned to one of the following household racial categories:

African American – all adults are in the African American group; OR a mix of adults in the African American and White groups.

Asian – all adults are in the Asian group; OR a mix of adults in the Asian and White groups.

Hispanic – all adults are in the Hispanic group; OR a mix of adults in the Hispanic and White groups.

Missing – any or all adults are in the Missing group.

Non-Hispanic White – all adults are in the White group.

Other – all adults are in the American Indian or Alaska Native, Native Hawaiian or Pacific Islander.

See this technical note on how we defined household race.

Telework Patterns

Telework was a hot topic in the past year, since there has been a necessary surge in the behavior, so we looked into what the data told us about pre-COVID telework behavior. The 2019 household travel survey asked, “How much did you work at home or telework for pay on your travel date? Here’s some things we found:

10% of workers teleworked between 6 and 12 hours a day. More workers teleworked (14%) less the six hours but more than zero.

People who teleworked part-time traveled 10% fewer miles by car during their average day; people who teleworked full time traveled 34% less distance by car.

The number of trips people who teleworked made was similar to those who did not. They had shorter trips in general, that is why their car distance is reduced. People substituted non-work trips for work trips.

Young people (18-24 years old) teleworked full-time less frequently compared with other age groups.

For the people with the household income of $150,000 or more, the odds of being more likely to teleworked were 30% higher than other income categories.

There is no difference in genders between the share of workers who teleworked. (More men work, however, so males represent a higher share of teleworkers [63%].)

Hispanic and Asian workers were 20% to 30% higher than Whites to telework. African American workers were 25% less likely to telework than Whites.

Home deliveries

As the number of home deliveries has appeared to increase dramatically over the past few years, we were interested to see what the travel survey data could tell us. The travel survey asked if households received packages or were delivered groceries on a given day. We found that in 2019, home deliveries were common and increasing. One out of three households received a package on an average weekday.

My daughter getting a delivery from Gram.

The Seattle Times also reported on the travel survey home deliveries data. The following charts show the trend in deliveries and that high income households received many more deliveries than low income ones.

Household Travel Survey Deliveries 2017 and 2019

Household Travel Survey Deliveries by Income

New mobility

Pre-COVID, new mobility (also known as ride-hailing and carsharing) was a hot topic. I’m sure it will come back.

We found that from 2017 to 2019, the share of regional adults who had ever used a ride-hailing service for travel went up from 24% to 35%. The share of Seattle adults who had used a ride-hailing service is substantially higher than the region as a whole. From 2017 to 2019, the share of Seattle adults who had ever used a ride-hailing service had increased from from 44% to 55%. In other words, over half of Seattle adults had used ride-hailing in 2019.

People who use ride-hailing and car share services have exhibit different travel behavior than the general Puget Sound population. Around 73% of ride-hailing and carsharing users commuted by walking/biking, carpool, or transit. In comparison, more than 70% of regional workers use a single-occupancy vehicle (SOV) on their commute.

Commuting Behavior

Transportation planners are often interested in commuting behavior. Our survey data helps to break common misconceptions around commuting.

For example, trips to or from work are only 24% of all weekday trips.

We’ve also heard speculation that lower income workers have to travel further than higher income workers to get to work as housing prices rise. The data shows, however, that:

From workers with incomes less than $75,000 annually, lower income workers have shorter commutes.Workers with incomes under $10,000 have commutes of around 14 minutes, whereas workers with incomes from $50,000-$74,999 have commutes around 27 minutes. The average travel time to work is about the same for workers in households with incomes greater than $75,000 – it’s around 33 minutes.

Model Estimation

After a while of working with the data, we realized that looking at data on one or two dimensions really doesn’t tell the whole story. We wanted to be able control for multiple factors to determine what variables had the strongest correlation with different behaviors. We estimated models to find what factors are driving behaviors. Some of code to estimate models is up here on Github.

This graphic shows the correlation matrix of variables in the auto ownership model. We use R for model estimation.

Source: Mary Richards, auto ownership estimation

The models looked at questions like:

Do people who telework on a given day travel less? (Yes)
Do people who live in places with pro-growth policies experience less residential displacement? (Yes)
Do people who get home deliveries make fewer shopping trips on an average weekday? (Not that we could observe)
Does transit access correlate with lower houeshold vehicle ownership, all else being equal (controlling for income, household size, etc)? (Yes definitely)

Let’s do this again sometime!

I hope we can form more working groups to build pathways from raw data through statistics to urban planning meaning. Data people tend to be introverts and often want to stay in their data corner by themselves. But when we work together our code improves, our statistical analysis has more rigor, and we are more useful to the planning process.

Why I like R

Tue, 22 Sep 2020 00:00:00 -0700

In the data science world there has often been a debate on whether R or Python is the better language. The debate may be less heated now given that recent technologies have allowed better integration of both languages within teams or projects.

If we are thinking of scripting in general, either language can automate tasks. Whether you use R or Python, you’re going to increase efficiency and replicability. Both have libraries to create dashboards, apps, and other visual and formatting needs as well. When you are learning how to script, either language will do.

I code in both R and Python and see the benefits of both, so why do I still describe myself as an R enthusiast and advocate? Here are reasons from my own experience:

Learning resources:

I wasn’t born a programmer and had a lot of false starts with programming, but when I started to learn R, it clicked faster than when I tried learning other languages. Partially it’s because I’ve found that accessing help for R was easier, and interpreting instructions and documentation was more user-friendly. Information about R and the documentation about different libraries are visually appealing and digestible. In the R universe, I was able to help myself easier than in other language communities.

Take RStudio’s Shiny gallery for example: https://shiny.rstudio.com/gallery/ (scroll down to ‘Start simple’ section). I learned how to build apps using the examples on their website. The way they’ve designed these examples was invaluable. I was able to replicate their demos and learn because I can see how things are produced in a clear and concise way. The mix of design, layout, and mediums of documentation itself made a difference in learning.

Lastly, my ability and interest to code in R is also because of in-house experts like Hana who is a great teacher and enjoys teaching others.

Tools:

RStudio’s popular Interactive Development Environment (IDE), a tool used to write and run scripts, is really well designed. This removes a barrier when trying to script. It’s intuitive and sophisticated but not overwhelming.

I’m a visual person, so having an IDE that easily displays plots, maps, or tables (that are also aesthetically pleasing) is a plus. Objects that I’ve created are listed in the IDE and viewable in a formatted view unlike when we print to the console. I can see what they look like under-the-hood and understand how the data is structured. This type of transparency of the data is effective in helping resolve hiccups quicker.

The ease of use in an IDE can make a difference in how I learn and plays a role that can either discourage or encourage me from coding. When I’m already struggling with the language syntax and the ins-and-outs of libraries, I don’t want another obstacle – figuring out how to use the software!

Visualization and document output:

R is a leader in visualization. By learning R, you don’t need to learn the trifecta of web languages that is JavaScript, HTML, & CSS, to build an app or website. Granted, learning R won’t solve all your website design problems or meet all your needs, but it definitely gives you ahead start in building a polished interactive web application.

R has great libraries to help you automate the production of documents (PDF, Word, Powerpoints). If your data changes, you don’t have to scour an entire document to update all your statistics. The code will do that.

I’ve heard a lot of comments from folks who want to have more creative control in their visuals, but are limited to what Excel or Infogram provides. Coding and learning R in particular, can really open up the visual and formatting possibilities.

Community:

My experience with the R community has been positive and encouraging. This stems from the R users and teachers who’ve helped me in my coding journey. This also stems from what I’ve seen organizations in R do, and the actions they take to make it a positive and inclusive culture.

RStudio is a certified B corporation that develops open-source software for R users. They have a substantial presence in the R community and over the last handful of years I feel like they’ve grown in terms of their outreach. They’ve created an annual conference dedicated to teaching and sharing resources. This includes livestreaming the conference and uploading recordings (accessibility!). They’ve engaged coders through contests, are hiring educators from diverse backgrounds/industries to teach and create practical tools and continually produce accessible webinars.

There’s also a theme of inclusivity, during the 2nd annual RStudio Conference, the Chief Data Scientist emphasized the “Pac-Man” rule: when gathering in a circle (chatting or in deep discussion) always leave an opening for others to join in.

I was also wary of the tech-bro culture and luckily have not experienced that in the R community yet.

It’s never too late to start coding! In an era of big data where everything is data-driven and our technologies are storing more and more data, coding is essential. I find that R is always innovating itself –evolving and changing (maybe at a faster pace than some might like), but it is so that we as users of that technology can do our jobs easier. My takeaway with coding in general is that embracing this kind of technology really helped me to help others and myself. So the last 4 years since I started learning R (and actually coding) has been a real game changer in how I do my work and how I’ve connected with my colleagues and surrounding communities.

So those are my primary reasons why I really like R. I also code in Python and I see some of the pros and cons to each. There are reasons for using one over the other depending on what you want to achieve, but I think if you’re looking for a gateway language, give R a chance.

Why are people displaced from their homes? Correlations abound, but prediction is never easy.

Wed, 17 Jun 2020 00:00:00 -0700

In the Seattle region, sky-rocketing housing prices, a legacy of discriminatory housing policies, and new transportation infrastructure, have all coalesced together to have the potential to displace people from their homes.

The Puget Sound Regional Council (PSRC) and the City of Seattle have been trying to understand displacement risk in our region. PSRC has developed a displacement risk index map to better identify places where people are at heightened risks for displacement.

Puget Sound Regional Council Displacement Risk Mapping

In PSRC’s 2019 household travel survey, residents were asked the reasons they left their previous residence. Data analysts at the City of Seattle and PSRC hope to better understand where displacement is occurring in the region using this data.

Polina, Christy, and I are some of members of the team analyzing the travel survey data on displacement. Pictures of people are always good for blog posts, especially data experts who are ladies.

Source: Picture during a Team Meeting

At PSRC, we have a team of analysts digging deep into this question of who is displaced. As we began looking into the relationships between single variables such as race, age, employment status, or income with displacement, it seemed to make sense to examine potential contributors to displacement more holistically within a model. The PSRC data team certainly doesn’t have all the potential predictive variables for displacement, but we have a lot of data lying around at PSRC, so I thought I’d take a shot and see what I could mine out of the data we have.

I chose to estimate a logistic regression model to identify which characteristics were most explanatory in determining a household’s displacement. The 2019 travel survey asked residents who had moved within the past five years the reasons they had moved. Only one member of the household would answer this question. Households which selected one or more of the following responses were labeled as displaced, and otherwise they were labelled as not displaced.

Housing Cost: Could no longer afford previous residence because of an increase in rent or
Income: Could no longer afford previous residence because of a change in income or finances
Community: Friends, family, or cultural community were leaving the area
Forced: Forced to move out (e.g., building demolished or renovated, asked to leave by landlord, foreclosure)

Displacement Estimation

The R library glm was used for the estimation. The code I used for model estimation can be found here. I’ve used many estimation packages in my day but I found the R glm library to be very friendly. One thing I struggled with during estimation is that I love using python pandas to manipulate data and I’m sorry, R, but dplyr and data.table just can’t compare for me.

Table. Residential Displacement Logistic Regression Results

term	estimate	std.error	statistic	p.value
(Intercept)	-2.28	0.37	-6.1	0
Household Income Under $25,000	1.7	0.18	9.45	0
Household Income is in the range $25,000-$99,999	1.04	0.11	9.57	0
Household Has less Cars than Adults	0.98	0.29	3.41	0.000642
Previous Home was Rental	0.73	0.13	5.6	0
LN(Distance to Light Rail from previous home)	-0.09	0.03	-3.16	0.001564
Household contains at least one non-Hispanic White Member	-0.31	0.11	-2.73	0.006379
Previous Home was Located in Seattle	-0.93	0.18	-5.14	3.00E-07

The model indicates that greater displacement is associated with (in descending order): household income, car ownership, renting as opposing to owning, being all people of color, living closer distance to light rail, and living outside Seattle.

I tried many land use and travel variables in the model, as shown above, but the only one that was significant was the distance to light rail. As expected, the income variables were the most explanatory. I was somewhat surprised to see the magnitude and significance of the variable about being located in Seattle. Seattle has strong renter protection laws than other places in the region.

The variable related to race was a tricky one to construct. How can you express the race of a household succinctly with limited data? The data is very thin for African-American households unfortunately. So I tried various aggregations of person racial categories such as Asian Only, non-Hispanic White Only, and African-American, Hispanic, Multiracial and other. Then somehow, I had to group the households into buckets, which wasn’t easy either because of the diversity within households as well. The variable that I ended up finding was significant was if a household contains at least one non-Hispanic white member, then the household is less likely to be displaced. I’d be interested to hear people’s thoughts on why this variable formulation worked best.

The pseudo-R-squared (McFadden, 0.10) of the model may indicate the variables in the model have limited explanatory power. A lower pseudo R-squared can be okay (as compared to R-squared) for explaining variable relationships to outcomes, as we are here, but I still wonder what isn’t being accounted for in the model. Then I started wondering if there was a better goodness-of-fit measure for a logistic regression. The self-proclaimed “Stats Geek” said that the Hosmer-Lemeshow goodness of fit test is appropriate for logistic regressions. The Hosmer-Lemeshow goodness of fit test resulted in the following measures: Apparently, a well-specified model has a p-value greater than 0.05, and the p-value for the the Hosmer-Lemeshow test was 0.89 so it looks like the model is well-specified.

Give me some ideas on this model!

Anybody got ideas on how to improve this model? How could I use it in the planning process? What variables are we missing? Do we just need more data? Is it problematic that how we have defined displacement? Is the model form itself a problem?

New Tool Release - the Household Travel Survey Data Explorer app

Thu, 21 May 2020 00:00:00 -0700

We have some exciting news to share! PSRC has just released Version 1.0 of the Household Travel Survey Data Explorer app.

Almost a year ago, we held a soft launch of the Beta version that allowed users to explore summary results from the 2017 Regional Household Travel Survey. Since the Beta release, we have been working on improving the details of the graphical user interface, integrating the app with new data from the 2019 survey, and optimizing the app’s performance upon loading.

It was an “achievement unlocked” moment when the app initially deployed as one of our goals was to make survey results easily accessible and efficiently sharable. A tool to crunch the numbers that would allow us to explore more easily what the survey contains and derive data-driven stories are other factors towards pursuing this project.

To make it possible, we developed (and are still developing) our technical prowess by learning R to harness its data wrangling and visualization capabilities. The app was written in R, powered by Shiny and other libraries such as data.table , ggplot2, DT, and Plotly for R to name a few. The process also involved designing a database schema to house iterations of the household survey and have the app read-in data from our SQL server data warehouse.

We’re continuing to fine-tune the app by addressing/resolving bugs and conceptualizing new features. All the while, we are continually quality checking results to make for a better tool and user experience. Overtime you may see some changes as we build up to the 2021 survey.

In the meantime, Version 1.0 of the app reflects the combined 2017 and 2019 Regional Household Travel Survey data. Generating data summaries of general statistics (shares and total estimates, sample counts, margin of errors) across 90-plus aspects (and cross-tabulation of aspects) is only a couple clicks away.

Happy exploring!

http://dataexplorer.psrc.org/household-travel-survey/

Biking in Seattle during COVID-19 time.

Tue, 05 May 2020 00:00:00 -0700

In the past couple of months, governments around the world implemented social distancing measures to slow the spread of COVID-19. During this unique time, we would like to understand how people’s travel behavior is changing. The first data point we wanted to analyze was the number of people bicycling for a couple of reasons:

The data is publicly available on the Seattle DOT website (link here);
Usually, May is the Bike Everywhere Challenge month in the PSRC region and Washington State (not this year, but we are hoping to participate next year);
Biking is one of the popular modes of getting to and within Seattle downtown.

Here are the questions we explored :

1. Have numbers of people biking decreased in March and April 2020 comparing to previous months/years?

2. Was there a decrease in biking after Work from Home or stay-at-home orders?

3. How has the people’s biking behavior changed?

Data

Seattle DOT has 12 bike counters providing bicycle and pedestrian counts on multi-use trails, streets, and greenways. The counts are available starting February 2014 to present, and most counts are updated once a month. Some of the trails don’t have March 2020 counts, hence in our analysis we wanted to focus on the trails that have March 2020 counts, including:

Fremont Bridge
Spokane Street Bridge
2nd Avenue at Marion St
2nd Avenue at Cedar St
Burke Gilman Trail north of NE 70th St Bike and Ped Counter
W 58th St Greenway at 22nd Ave NW Bike Counter
Elliott Bay Trail in Myrtle Edwards Park Bike Count
7th Avenue and Blanchard St
Westlake Avenue at Newton St
26th Avenue Greenway
MTS Trail west of I-90 Bridge Bicycle and Pedestrian Counter
Broadway Cycle Track North Of E Union St Bicycle Counter

Analysis

1. Have numbers of people biking decreased in March and April 2020 comparing to previous months/years?

To understand if the number of people biking changed, we’ve calculated a percentage change of bicycle counts in March and April (2019 vs 2020).

Trail Name	March 2019 vs 2020 Delta, %	April 2019 vs 2020 Delta, %
Broadway and E Union St	2.26	-20.2
Burke Gilman	-5.0	65.5
Elliot Bay	-21.8	20.4
26th Ave Greenway	-24.4	10.8
NW 58th Greenway	-30.9	-19.8
Fremont Bridge	-32.3	-25.7
Spokane Bridge	-33.2	-2.1
MTS Trail	-36.7	41.6
2nd Ave and Cedar	-46.4	No data available
2nd Ave and Marion	-50.2	-68.6

Note: the counter at 7th Avenue and Blanchard was not included because the counter started collecting data in August 2019. Westlake and Newton counter was also excluded because of data quality.

As expected, the bicycle counts declined on all trails/streets except for Broadway and E Union Street location in March 2020 when compared to March 2019. Some areas like Seattle Downtown and Belltown having a steeper decline (2nd Avenue and Marion and 2nd Avenue and Cedar) and some trails have a more shallow decline (Burke Gilman and Elliot Bay trails). In April, there is a different picture where recreational trails Burke Gilman, Elliot Bay, MTS trail, and 26th Avenue Greenway saw growth in bicycle counts when comparing to April 2019.

2. Was there a decrease in biking after Work from Home or stay-at-home orders?

On March 4th, the King County officials gave the recommendation to work from home to the companies, and government agencies and tech firms advised their employees to work from home. Stay-at-home order was placed by the Washington State on March 23rd. The next two figures show February-March 2020 bicycle counts for two different types counter locations: Downtown/Belltown (2nd Ave and Marion St; 2nd Ave and Cedar St) and trails (Burke Gilman and Elliot Bay trails). These four counters will represent commute biking routes and recreational routes respectively.

Bicycle counts in Downtown/Belltown started decreasing at the start of March, right after Washington State and King County’s work from home recommendation. The downward trend continued after the stay-at-home order. The bicycle count decline along commute-heavy routes was expected and not surprising.

The Burke Gilman and Elliot Bay trails didn’t see such a steep decline in bicycle counts. Moreover, the highest bicycle counts occurred after the work from home recommendations and went down after the stay-at-home order. An interesting shift occurred in bike counts between the Burke Gilman and the Elliot Bay trails. In February 2020, bike volumes were higher on the Elliot Bay trail than on the Burke Gilman. Generally, bike counts on Elliot Bay trail were higher during weekdays vs weekend - this suggests that this trail was mostly used for commuting in February 2020. In the middle of March 2020, Burke Gilman trail saw bike counts growth that exceeded bike count on Elliot Bay trail.

3. How has the people’s biking behavior changed?

To explore the behavior change question, we decided to look at the average day in April 2019 and April 2020 for the counters with available data for April 2020, which includes Burke Gilman, Elliot Bay, Fremont Bridge, and Spokane Bridge.

In April 2019, there are two peaks during a weekday: one in morning (around 7-8AM) and another one during evening (around 6PM). This pattern represented a typical commute for people when peaks were before/after 9am-5pm working schedule, and each peak is no longer than one hour. A typical weekend had one peak starting from 12pm to 3pm.

In April 2020, when the majority of people don’t have to travel to an office, weekdays’ patterns look quite different from April 2019. There is one pronounced peak around 6pm during a weekday in April 2020. The bicyclist might be talking biking trips or exercise trips after work. In general, there are not that many people who bike on a weekday compared to a weekend.

Results

The biking patterns have changed during COVID-19. At first, in March, the number of people bicycling collapsed along commute-oriented routes such as ones in Belltown and Downtown, while a decline in bicyclists was not that steep along Burke Gilman and Elliot Bay trails. In April, the number of bicyclists skyrocketed on Burke Gilman and Elliot Bay trails comparing to April 2019 number. The weather might be one factor - April was warmer with less precipitation this year.

Peoples’ biking behavior is changed as well - you will see more bikers during a weekend than on a weekday. There are no typical 8AM and 6PM peaks during the weekdays, but one 6PM peak.

Toward version 1.2: Improvements in SoundCast

Wed, 19 Oct 2016 00:00:00 -0700

This week we marked a milestone in the continuing development of our activity-based travel model, SoundCast. The latest official release, version 1.1, included big improvements to the truck model and the benefit-cost tool. Now, SoundCast “1.2 milestone 1” includes lots of new goodies.

Improvements in 1.2 milestone 1

Faster. Between code improvements and some new server hardware in the closet, we have significantly decreased runtime from 30 hours to 17 hours for a full run. This is mostly Moore’s Law in effect: incremental updates in computing power over the three years since our last hardware purchase are responsible for most of the improvement.

Bike trips are now assigned to the network. A completely new bike assignment algorithm is now implemented. The chosen bike routes are informed by auto volumes, amount of uphill climb, and presence of different bike facility types.

Initial results of brand new bike assignment model. You can see the Burke-Gilman!

Source: PSRC SoundCast

We also have many more bike counts now, with which to validate the results. This model is just a first effort, and more tuning will be coming in the near future.

Simplified Toll/No-Toll representation. After extensive review and testing by both PSRC and our consultant team RSG, the toll/no-toll choice has been turned off in SoundCast. This results in:

Fewer skim matrices
Fewer modes
Slight speed improvement

—without any significant loss in capabilities; as the choice of whether to choose a route containing a toll is still embedded in the EMME pathbuilder.

Big update to “external” trips. Trips that either originate or end outside of our four-county region are known as “external” trips, and those hadn’t been updated in a long, long time. This revision now bases the distribution of the external trips on actual observed behavior from LEHD flow data for work trips, which are about 70% of all external trips. We are still working on improvements to the other (nonwork) 30%.

Thousands of daily trips are 'externals'—trips which enter or exit our region.

Source: PSRC SoundCast

Model re-estimation based on 2014/2015 household survey. Many subcomponents of the travel model have be re-estimated using the latest survey data, including auto ownership, work tour and work-based subtour mode choice, school mode choice, other home-based, and more. Destination choice and day-pattern model re-estimations are still to come.

Can I use this?

This is just an internal milestone and as such, is not fully tested. You’re welcome to explore the code at GitHub or download tag v1.2m1 locally—everything SoundCast is fully open-source. The official release of SoundCast 1.2 will occur later this winter, after the model re-estimation, calibration, and validation is complete.

PSRC Data Stories

Shiny Data Visualization for Urban Planning

Why do we use Shiny to visualize urban planning data?

Top Seven PSRC Shiny apps

Click on the links to see the apps live.

Hoping to inspire others!

Shiny For Planning

Shiny Data Visualization for Urban Planning

Suzanne Childress

2024-10-31

Shiny for Urban Planning Data

Conclusions

Women's transportation needs

Women’s Diverse Travel Needs Often Go Overlooked

The Puget Sound Regional Council Household Travel Survey provides data on travel by women

Travel demands put different pressures on women

Gender and race intersect in transportation needs

Women of color use transit most

Women bike far less than men

Regional and nationwide data show similar constraints

Changes brought about by COVID-19 pandemic

Women of color continued to use transit more during COVID-19

Trip purposes by gender changed from 2017/2019 to 2021

Conclusion

The Diversity of Asian and Pacific Islander Experiences

Across Heritages and Where People Live

And the challenges but importance of racial equity data

How do worker earnings for Asian and Native Hawaiian Pacific Islander people compare to other groups?

How do worker earnings vary across different groups of Asian, Native Hawaiian, and Pacific Islander people?

Intersecting diverse heritage and geography

Using Statistical Evidence to Explain Experiences of Smaller Populations

Discovering Household Travel Data Stories Through Collaboration

Thank you to Abdi Mohamud, Brian Lee, Christy Lam, Drew Hanson, Grant Gibson, Neil Kilgren, Polina Butrina, and Mary Richards who shared their brains and time in this effort!

Upskilling Before Discovering

Data Discoveries

Residential Displacement

Understanding travel by different racial groups

Telework Patterns

Home deliveries

New mobility

Commuting Behavior

Model Estimation

Let’s do this again sometime!

Why I like R

Learning resources:

Tools:

Visualization and document output:

Community:

Why are people displaced from their homes? Correlations abound, but prediction is never easy.

Displacement-Related Reasons for Moving

Displacement Estimation

Give me some ideas on this model!

New Tool Release - the Household Travel Survey Data Explorer app

Biking in Seattle during COVID-19 time.

Data

Analysis

1. Have numbers of people biking decreased in March and April 2020 comparing to previous months/years?

2. Was there a decrease in biking after Work from Home or stay-at-home orders?

3. How has the people’s biking behavior changed?

Results

Toward version 1.2: Improvements in SoundCast

Improvements in 1.2 milestone 1

Can I use this?