Stats Analysis Shows 20-30% Gains Possible

I have a strong belief that the potential for PE is a percentage of your existing size, based on the idea that each cell has a “stretch”/cell division potential

I havn’t seen the PCT approach on the data on the site (maybe I havn’t searched enough), therefore I would like to share.

Therefore I wanted to examine the data based on how much percentage people grew and what effort was needed.

This has resulted in the attached scatterplot.

Looking at the graph to me the results are:

~20% gain is doable with the right effort, but also a little luck (a lot of people in +1000 days with <20% increase)

~30% gain requires luck and/or effort.

It also compares with the clinical test of the extenders.. 20-30% gain in 6-12 months, after that it is diminishing returns (though probably you can get some results with the right effort).

Just want to be clear on your graph:

Y axis is percentage change
X axis is data count (number of entries) or is it days of PE?

Nice work.

What are the axes of the graph? Are you plotting BPEL and MSEG?

Have you compared this to a straight gains scatter plot with the same sample restrictions? There’s one on the stats analysis page and if you chop out the negative side and restrict the sample more, they don’t look markedly different.

Do you think this helps confirm your percentage thesis?

BPEL Gains in Percentage (Y axis) and Days from first to last entry (X axis).
I’ve done a new plot with labels on the axis (and included trendlines, see below).

A lot of further data adjustments can be made.. I already sorted out some of the most “incredible” gains.. Like 800%.!
Also I removed series that only consisted of 14 or less days of reporting.

I also examined the average and std deviation:

Total set: Average is 8,42% gain and std. Deviation of 11,9
Total set of +60days exercise: Average is 10,92% gain and std. Deviation of 8,39
Total set of +120day exercise: Average is 12,48% gain and std. Deviation of 6,18

So according to my fast detoriating statistical knowledge 95% of data should be <25% gain at +120 days of exercise.
I’ve also put in two trendlines.. A liniear and a log one. I have no idea which one is the right fit (I know you can F-test this or something), but I believe it is actually somewhere in between.. Ie. You will experience the law of diminishing returns.. But you will very slowly if you are consistent add to your max potential. Ie. A a*x + b*log(x) something function.. But to add this trendline requires more than my knowledge (or more correct, it’s beyond what I have time for currently).
To do this right you should probably fit the function to each data set (username) and aquire the “a” and “b” and do statistics on that.

Not sure about using standard deviation on percentage numbers.. Cannot remember my statistics well enough.
I do have some (formal) statistical knowledge .. But not a degree in such.. So somebody with a real degree in statistics should maybe look at it.

And remember.. All this was just done to satisfy my own curiosity, but I thought you would like to see the data too :-)

AND remember this is data analysis on the data set.
This means that this data should NOT be viewed as the definitive potential.. It should be viewed as the potential of the average user who has average resource (of will-power and time) to commit to the purpose.

Ie. If you are really really (above average) dedicated and committed and have the resources (mainly I believe will power .. And maybe luck) you will probably be in the above range of the set, ie. 30-40%.

I’m interested in the outliers around 50%. Did you screen larger gains or is that the probable max?

I think you are right about the trend being somewhere between.

How to analyze the data with Excel the way I did (reproduce my results). Requires some Excel knowledge:

1. Get the .csv data file and import it (if you cannot do this yourself .. Stop here)
2. Mark the whole dataset including the top labels and chose the menu “insert” and click the “Pivot table” and put it in a “new worksheet”
3. In the pivot table you should start by dragging the “username” down into the “rows” PivotTable Fields. The first row should now be the usernames.
4. Now drag the “BPEL” down into the “values” of PivotTable fields. By default it will be a sum (which makes no sense)
5. Right click on the label “Sum of BPEL (cm)” in the pivottable and chose “Summarize Values by” -> “Min.”
6. Do step 4+5 again, chose instead “Summarize Values by” ->”Max.” and get a new column of “Max. Of BPEL(cm)” so now you have both a “Min.” and “Max.” column for each user.

Now you have the raw data to calculate progress on. You can do this with standard excel formulas (ie. Put “=C4-B4” into D4 and expand this to the complete column)
You can do this for the date as well to get the days difference between the entries.
On this data I’ve done some filtering.. Ie. Take out data that looks wrong (like +800% gain) and data with only one recording etc.
I now just did the formulas for percentage gains and scatter plotted it.

You can do the above for all the other data-entries in the set. I was only interested in BPEL.

Here’s a histogram regarding the percentage BPEL gain.
The bins are in 5% intervals. Ie. 0-5%, 5-10% and so on.
It includes a pareto-line which will tell you how many of the total population is less or equal to the given bin.

Looking at the chart, I would guess it actually consists of two data-sets.
“Normal” and “Hardcore”.
You see the same thing in timetables for public sports events, like long distance running and the like, ie. The dataset consist of two bell-curves joined together (average Joe-the-hobby-runner and professional runners).
It is not as visible but it could look like it and maybe not. I just think the bin (20,25] and upwards looks a little strange, I don’t know.
(maybe I just want it to be so)