In this second post in my series "Exploring my physical data", I take my first steps in exploring my own physical activity data.
The first hurdle in my exploration is gaining access to my own physical activity data. I don't use any smart wearables or actively track my own activity, but I've allowed Google Fit to passively track my physical activity since 2018.
Thankfully, Google has its own service, Google Takeout, that allows you to access and download pretty much all of the data they have on you in a simple and easy to use interface.
I used this guide to quickly find out how to export my Fit data, but Google Support has a really useful guide as well.
I decided to only use the daily aggregated data (ie daily steps, daily distance walked, etc.) as the info on individual activities seems a bit like overkill at this stage.
You can select all sorts of different settings and parameters. I opted for a csv file of the daily aggregated data. I then added columns for day of the week, month and year with Excel to make it easier to work with.
I then quickly created graphs for each of the main parameters by date for two reasons: 1. I was eager to start exploring straight away; 2. to check whether any of the parameters seemed unreliable and should be excluded from any further exploration. I found that most of the parameters tracked very closely (as they should), but the speed data seemed quite variable and often well out of the realms of physical possibility!
For example, when Usain Bolt clocked his World Record in the 100 m sprint his top speed was 27.8 mph. That translates to a little over 12.4 m/s, or really damn fast! But, according to my Google Fit data, I make Usain Bolt look like a total chump. Google, or my phone, logged me exceeding Bolt's World Record pace a total of 85 times in 2019, with one day reaching a max of 205 m/s (458 mph)!
A quick search explains this very simply, 458 mph is around the cruising speed of a jet plane, which makes sense as I was flying to Zurich on that day.
The problem is that Google speed data is fed by GPS data and not accelerometer data. This reliance on GPS means that there is a risk that periods without signal will show up in the records as huge values. If you take the 205 m/s example, there are no corresponding blips the next day, when I got the train to Basel, or the day after that, when I flew back home.
With that in mind, I decided to focus only on parameters that were measured using the accelerometer, or a mixture of GPS and accelerometer data.
While step detection and the like are still not perfect, it doesn't seem as likely as the speed data to give completely wild values (I logged 4135 steps and 52 active minutes on the day I hit 205 m/s).
The major problem with this data set is that it relies on me having my phone on and with me at all times. As a typical millennial, this happens most of the time but there are occasions where I will leave my phone behind and work in the garden or something. I'm not overly concerned about this issue though, as I'm really only looking for general trends and there won't be much value in looking too closely at individual data points.
Next time, I'm going to start playing with these data and my new friend, Tableau.
The first hurdle in my exploration is gaining access to my own physical activity data. I don't use any smart wearables or actively track my own activity, but I've allowed Google Fit to passively track my physical activity since 2018.
 |
| Google Fit, the source of all my physical activity data |
Thankfully, Google has its own service, Google Takeout, that allows you to access and download pretty much all of the data they have on you in a simple and easy to use interface.
I used this guide to quickly find out how to export my Fit data, but Google Support has a really useful guide as well.
I decided to only use the daily aggregated data (ie daily steps, daily distance walked, etc.) as the info on individual activities seems a bit like overkill at this stage.
 |
| You can download your own physical activity data and have a look for yourself at takeout.google.com |
You can select all sorts of different settings and parameters. I opted for a csv file of the daily aggregated data. I then added columns for day of the week, month and year with Excel to make it easier to work with.
 |
| A snapshot of my exported physical activity data |
I then quickly created graphs for each of the main parameters by date for two reasons: 1. I was eager to start exploring straight away; 2. to check whether any of the parameters seemed unreliable and should be excluded from any further exploration. I found that most of the parameters tracked very closely (as they should), but the speed data seemed quite variable and often well out of the realms of physical possibility!
For example, when Usain Bolt clocked his World Record in the 100 m sprint his top speed was 27.8 mph. That translates to a little over 12.4 m/s, or really damn fast! But, according to my Google Fit data, I make Usain Bolt look like a total chump. Google, or my phone, logged me exceeding Bolt's World Record pace a total of 85 times in 2019, with one day reaching a max of 205 m/s (458 mph)!
A quick search explains this very simply, 458 mph is around the cruising speed of a jet plane, which makes sense as I was flying to Zurich on that day.
The problem is that Google speed data is fed by GPS data and not accelerometer data. This reliance on GPS means that there is a risk that periods without signal will show up in the records as huge values. If you take the 205 m/s example, there are no corresponding blips the next day, when I got the train to Basel, or the day after that, when I flew back home.
With that in mind, I decided to focus only on parameters that were measured using the accelerometer, or a mixture of GPS and accelerometer data.
 |
| According to these data, I really need to change my day job! - Graph created in Tableau Public |
While step detection and the like are still not perfect, it doesn't seem as likely as the speed data to give completely wild values (I logged 4135 steps and 52 active minutes on the day I hit 205 m/s).
The major problem with this data set is that it relies on me having my phone on and with me at all times. As a typical millennial, this happens most of the time but there are occasions where I will leave my phone behind and work in the garden or something. I'm not overly concerned about this issue though, as I'm really only looking for general trends and there won't be much value in looking too closely at individual data points.
Next time, I'm going to start playing with these data and my new friend, Tableau.
Comments
Post a Comment