This is a self-derived and motivational research initiated to test statistics and see how data mining, visualization, and predictive analytic is not a bogyman. Have fun!
Weight loss is a crucial issue in the United States today. Research shows that excessive weight increase our risk of heart and other diseases. The Body Mass Index (BMI) is one way an individual effectively monitor his or her weight to police their weight in order to decrease their risk of illnesses. Even though illnesses are caused by many factors and not just physiological and biological characteristics, the condition of physical body is paramount to reduce or increase our chances of being sick.
You may be losing more weight than you think. How to determine that depends on your patience and motivation to track, mine, analyze, and build simple predictive models that could help tell a really cool story how to manage your weight, health and money.
Overview
Having said that, I personally undertook a personal research project to monitor my weight and see how I am physically doing and to see how I can loss some extra kilograms of bodily fat and all that’s included in the package.
Objective
To this end, I downloaded the Google Play Store App called Pedometer and started to monitor my daily moments (walking) to see how that facilitate the process of weight reduction and reduced BMI score.
Method
I will collect pedometric data for 8 months beginning August 2015 until April, 2016. Data will be collected with the pedometer app installed on my cellphone. Data are collected on a 24 hours basis. Data that will be collected include number of steps, kCal, date, time, distance cover in mile, the daily average speed in miles per hour.
Other data that are being collected include daily weight (kg), time of the day weight was recorded (usually early morning), Max Heart Rate, Heart Rate at Rest, Age ( a constant factor), Gender (dummy variable), frequency walked (i.e. the number of times a day walking was initiated, not the number of walks), and Days in Action (number of days walked).
At current, I have managed to collect pedometric data from August 2015 through November 4, 2015 and hopes to continue this for an additional 6 months. Data are not collected when I am asleep…duh:)
All you need to do is launch the app once on your cellphone and as you walk, data are generated. All you need to do is to have your phone on you and that means, in your bag, pocket, etc. No data will be collected if you leave it at your desk or your call, which I am a usual victim off. So, the more accurate data you can collect depends on your commitment to have your phone on you all the time. Also, no data will be collected if your phone battery dies. It has to be charged at all times to be able to collect data. At the end of the 24 hour, a new data sheet/template starts immediately and it is auto-correlated with time.
For my case, I usually take my phone when I am about to leave for work during weekdays and weekends as I will see fit. I put the phone away when I am ready to take night shower and that’s it.
Analysis & Anticipated Outcome
The data collected in this self-initiated study will be analyzed using advanced statistical techniques in SAS JMP Pro 12 with the goal of building a self-derived predictive model that can be customized to any individual situation or circumstances for the purposes of weight loss and BMI score normalization and or management or just to have some statistical fun with real world stuff.
Preliminary Analysis & Result
Figure 1: Graph Builder shows the predicted effects of Days in Action, # of Steps taken, #kCal burned, and Distance covered on Weight (kg) and BMI. Click on the image to view the stats.
The cubic transformation of the number of days in action seems to have decreasing effects on weight loss and BMI. As the number of days in action increases, the overall BMI and bodily weight loss decreases proportionally relative to their increasing root square values (0.85).
However, #steps seems to be fairly variable on a daily basis, but show symmetrical relationships between the predicted response. Though the number of steps changes on a daily basis, bodily weight loss and BMI scores tend to decrease when more steps are taken and increase when fewer steps are initiated.
Similar preliminary visual association can be deduced from the graph relative to #kCal and Distance (mile). This is partially due to the smallness of the data, but is expected to change as more data are collected and automatically re-run this analysis.
The essence of this visual analysis is to identify relationships that exist between or among predictor variables and how those predictor variables impact the response (predicted) variables.
More to follow:)
Here is a snapshot of the pedometric data table. Click on the image to view the data. Some data are derived variables, like time in action.
