[Step 2]: Starting to explore my own physical activity data

Welcome to an exploration of my own physical activity data form 2019, captured by Google Fit. 

This is Step 2 in a multi-part blog series in which I take my first steps into data visualization by analysing my own physical activity data recording by Google Fit. In Step 1, I downloaded my own data and highlighted some oddities about how Google records physical activity data. You can see the first post here. 

You can also see the full data set, as well as interactive versions of these graphs on my Tableau Public profile. 

First, let's look at the daily step count across the whole time period, July 2018–January 2020. 

My daily step count from July 2018–January 2020

Just looking at this graph shows me quite a few things, even without any analysis or looking individual data points. Most of the major spikes in activity are related to holidays or travel - the spikes in July, August and November are all trips away from home. You can also clearly see the expected drop in activity seen around Christmas 2018, but not in 2019. This difference in activity over the Christmas break is most likely explained by us gaining our lovely dog, Skye, in March 2019. 

Even without looking at the data in great detail, you can see that when we rescued Skye in March there was a sustained increase in daily steps. This increase is because we spent a lot of time walking her to build up a bond and to try and give her lots of exercise, so she was tired and more likely to sleep through the night. By April, you can see that we've eased off on the exercise, this is in part because she was getting more settled but also because we learned that excessive exercise can actually lead to increased anxiety in sight hounds. After that, we got into a good routine and it looks like most days are hitting at least 4,000 steps (apart from a few lazier Sundays!). 

This is Skye, she's a 3-year-old rescue (we think Saluki-Greyhound cross)

So what can we learn about these data? I decided to look at just 2019 for now, so we're working with a full calendar year. I'll look at the full data set in combination with some other data sets in later posts. I also haven't excluded any data from when I was on holiday or when I was sick, as this is just an initial exploration of the full data. First, let's look at the distribution of daily step totals to see what the most common daily step totals are. 

In 2019, my mean daily step count for the whole year was 7,060 steps, while the median was 6,886 steps. As you'll see from the histogram below, the mean is likely skewed by a few extreme values. You can see that the majority of daily step totals fall between 4,000–11,000 steps. 

Histogram of daily steps in 2019

If we look at these data in another way, you can see the distribution is skewed fairly heavily, with 7 outlying values above the upper whisker of the box plot (upper whisker contains data points within 1.5x the interquartile range). You can also see that 50% of the data (25–75th percentiles) fall between approx. 5,000 to 9,000 steps. 

Box and whisker plot of daily steps in 2019. Boxes indicate 25–75th percentiles, whiskers are calculated as 1.5x IQR. 

So what do these data tell me about my own physical activity levels, am I an active person? 

A 2017 global study, published in Nature, found the average value for UK participants was 5,444 daily steps. So my median 6,886 steps seems pretty good. In fact, 69% of my daily step counts in 2019 were above the UK average. 

However, the most common  target is 10,000 daily steps. While the exact benefits of hitting 10,000 steps over any other target is not clear, it seems like a reasonable amount of activity and many people use it as a target. In 2019, I only hit 10,000 steps or more on 63 occasions, or 17% of the time. So, while my daily activity is above average for the UK, I could probably only claim to have been moderately active in 2019. This doesn't capture any activities I might have done without a phone in my pocket (such as workouts, swimming and badminton), but I never managed to do these activities consistently enough for me to claim to be anything more than moderately active. 

I also looked at my minutes of daily activity, which paints a similar picture. The general consensus is that you should get a minimum of 60 minutes physical activity per day. However, physical activity comes in all shapes and sizes; for example, the WHO recommendations say that people aged 18-64 years should do at least 150 minutes of moderate-intensity physical activity throughout the week, or at least 75 minutes of vigorous-intensity physical activity throughout the week. My mean daily active time was 74 minutes per day, without counting for other sports activities not tracked by my phone and doesn't give an indication of how vigorous the activity was. 

If we look at the data on a month-by-month basis, you can see how much my daily activity varies. Here I've shown daily step count, but the active time data is very similar (you can see them side by side in my Tableau Public profile).

Distribution of my daily step counts for 2019, grouped by month 

Looking at this graph raises a number of questions that I'd like to try and answer in the next posts. It appears that Spring/Summer months might have a higher average daily step count than Autumn/Winter months, but there's a lot of overlap between any given month. I'd like to explore whether there is a relationship between daylight hours, weather, average temperatures and daily activity levels. It also appears that my daily activity appears to be higher when I'm on holiday, which I'd like to explore further to see if this applies to any time that I'm away from home.

The final visualisation of these data I'll show you for today is the distribution of data by weekday. I wanted to see whether there were any obvious differences between weekdays and weekends, and the differences are quite clear. 

Distribution of my daily step counts by weekday in 2019

You can clearly see that the range of values is much wider on weekends than during the week. This rings true to my own experience, some weekends were very active, while others were much more relaxed. It seems I had more lazy weekends than active weekends, as the median values for the weekend are sitting below the 25th percentile for weekdays. This also shows me that I have fairly active and consistent weekdays - I get the train to work, so there's an element of walking to and from the station between home and work. I'd be interested in seeing whether the separation between weekday and weekend is made stronger if I excluded non-typical days (holiday, work travel, sick days) from the analysis. I'll look at this in the next post. 

So far, I've only been looking at simple descriptive statistics about these data. While I can use these graphs to to look back at my data for 2019, it doesn't help me all that much with how I could improve my daily activity this year, nor does it help me understand why my physical activity looks the way it does. For that, I need to use some more advanced statistics, as well as combining these data with other data sources to examine their relationships. 

In the next few posts, I'll start combining these data with other data sets to see how my data relates to other data sets available around the world. I'll also be looking at an updated, more informative data set to see how my physical activity is looking in 2020 and to explore how my activity levels have been affected by the COVID-19 pandemic.