#personalexperiments

[foxxiauddi]

So far so good....giving my little tater tot kisses before he goes to school and I was very surprised to see none of the lipstick come off. I may have been contemplating buyers remorse with this, but now I’m leaning towards the 5 star area...The test isn’t over yet, let’s see how well this stuff lasts for a workday. Comment or something if y’all wanna see more product review. #luxeory#lipstick #i like testing theories #personalexperiments #taterkisses

[ralphjpassarella]

Personal Experiment #2: Comparing precision and accuracy between multiple accelerometers – Protocol

My second personal experiment aims assess the (relative) utility of digital health activity trackers. This experiment isn’t an N-of-1 trial like our first installment. This is more of a classic measurement error / calibration experiment. So without further adieu, the design…

Aim: This experiment aims to assess the precision and accuracy of three commercially available activity trackers at measuring steps taken by a user of the device.

Experiment: I will wear three activity tracker devices, the Fitbit One, the Fitbug Orb, and the Jawbone UP, as directed by each manufacturer. For fourteen consecutive days I will wear the three devices, take 500 steps, and manually count each step using a commercially available step counter.

Prior to beginning the 500 steps, I will record the number of steps registered on each device for the day for a device’s baseline reading. After taking my 500th step, I will record the number of steps registered on each device. I will subtract the baseline step count from the final step count to arrive at the number of steps measured by each device during my 500 steps.

I will perform two trial runs of 100 steps each before the first experiment to familiarize myself with the process of uploading data from each device.

I will record measurements from the three devices being tested in the following order: 1) Jawbone UP, 2) Fitbit One, 3) Fitbug Orb

Primary outcome: Accuracy of each device compared to the “gold standard” measurement of self-recorded steps via step counter.

Secondary outcomes: Precision of each device over the course of the 14 trials. Relative accuracy of each device compared to the other two devices.

Analysis plan: [I’m going to write an updated post with the analysis plan before looking at the data. I figure I’ll make myself do it if I just get started with the experiment.]

Here is a link to the google spreadsheet where I’ll be collecting the data.

Potential sources of error:

- Additional steps counted by Fitbit, Fitbug, and/or UP either 1) before experimenter begins counting steps or 2) after 500 steps have been recorded but before readings are taken from a given device.

- Miscounting of steps by experimenter when operating the manual step counter – this could include over- or under-estimating total number of steps taken while measurements are being recorded.

- Error introduced by changing the terrain over which the 500 steps are taken

Materials:

- Step counter

- Fitbit One

- Fitbug Orb

- Jawbone UP

- Laptop with internet access

- iPhone to collect data from Fitbug device

[ralphjpassarella]

Personal experiment #1: Lunch purchasing intervention – Results

I’m not terribly certain how to write up the results of my N-of-1 personal experiment. I think the raw data is most compelling to me and is probably most intriguing to anyone reading this, so here they are in a google spreadsheet.

By way of a tally, I purchased lunch 9/20 days during the 20-day run-in period when I did not send myself a text message reminder, and I purchased lunch 10/20 days when I did send myself a text message reminder during the 20-day treatment period. The intervention did not appear to have a “significant” impact (I put “significant” in scare quotes to humor myself).

Following my two grocery purchases during the 8-week period, I went six days and four days without purchasing lunch respectively. This might seem obvious, but I found that having chips, particularly Sun Chips or Baked Lays, correlated with my not purchasing lunch (data not recorded). As I ran out of my favorite chips, I began purchasing lunch again.

I did not track the following variables as I had intended per my methods section: perceived hunger level, bank account status, cash in wallet, friends’ desires to eat lunch outside the office, weather

Other “supplementary results” include:

- My average wake time was 4:50

- My median wake time was 4:45

- My average self-reported stress level at lunchtime on days when I purchased lunch was 5.21. It was 5.23 on days when I did not purchase lunch. I’m paying no mind to significant digits if you’re wondering.

- My average self-reported workload on days when I purchased lunch was 6.68. It was 6.71 on days when I did not purchase lunch.