Size's PE data analysis

Try a correlation using the inter quartile ranges.

-This would remove data outliers correct? Are you sure this is the best way to remove the outliers? I know matlab has this cookie-cutter “smooth data” feature in its curve fitting toolkit that smooths the data nicely. Or I may just remove the outer quartiles as you suggest.

Then try it for different bands of starting LENGTH.

-Do you mean to split the girth data based on starting length? Or are you just saying to do the same things for starting length as you are proposing for girth?

My first guess is that the ultimate correlation while negative as you suggest is far less negative than that trend line. Also try it for guys who stuck with it the same amount of time (a bivariate correlation of gains against length of career and starting girth)

-Cool. I will look into this.

Maybe guys with big girth are less motivated because they have “big dicks” to start. So if you if you look at guys who gave it the same amount of time and effort as PE’ers that the gains in length will be smaller for thicker girth but much less than that line indicates, and your statement is a partially a psychological fact rather than a purely anatomical one. The true anatomical one would be more interesting.

-I agree. I think adding the length of PE career into the analysis should help.

Great stuff Tube, keep at it. The hell with that monkey jelqing… ;)

See what happens to R squared when you remove outer deciles, octiles, etc…

Then try to see what corellations between starting girth results within different sets of starting length bands. I think any level of outlier elimination will reduce your slope magnitude (it’s already small) and increase your R squared. Quartiles is probably too extreme. But right now the .04 is a non-analysis (no strength-linearity is not supported) and the .25 while not completely negligible is pretty small. What are the confidence intervals at 90% and 95% on the beta? Matlab should give those.

I think the most important thing to look at will be to use the “rate of gain” statistic rather than absolute gains. This indexes for time.

Thse are just the slices and dices that I would intuitively do just to see what happens. Although indexing for time spent pe-ing must be done. It wont matter if on average each group spent the same average time PEing but I doubt it. I think the thicker guys quit earlier partly because they started longer. What is the corellation for PE career length and starting length? Bet its negative and moderately significant. Then starting length and girth? Other studies have put this at .4

I’m hoping for some interesting surprises though.

I quickly spit out the 95% confidence points for the linear fit.

Values are
slope: -0.233 (-0.3555, -0.1106)
y-intercept: 1.946 (1.329, 2.562)

I included a graph showing the bounds and I expanded the perspective out to zero starting width to show the y-intercepts. It matches with the values above. This reminds me that I occasionally know what I am doing. ;)

Try regressing average gain rate— exp(ln((total gains)/(starting length))/(months of PE))— vs. Starting Girth

The confidence intervals show that the corrlation is negative, but the R squared shows that only 4% of the variation in gains is explained by this factor alone with the outliers included on an absolute basis.

Will MatLAb spit out a flat comma-delimited file of the data points? If so could you attache a zipped version to your next post?

About the R^2: With as much variance as the data has, I think 4% is interesting.

I will be removing outliers in due time and also doing the gain rate plot. It may take a few days. I’ll also take a closer look at your formula as it is not familiar to me at first glance.

I know in my engineering days we used to do this analysis that went like — Find the probability that a specific distribution would occur from some generating function, like a gaussian with a certain variance. I would need to review that though.

What data do you want in your CSV file? I will put together something. Give me a couple days. :)

Tube

In my formula substitute “ending length” for “total gains”, sorry, my error. The formula is just the computation of a gemoetric average monthly gain rate.

On the CSV, I did it myself, in Excel. I have to get Matlab.

Actually I think that the variance is mostly determined by data that is subjective or not included in Size’s input set - Things like “devotion” and “intensity” can’t really be measured but check boxes for things like:

used heat
multiple workouts per day
length of average workout
days on/off
hung, jelqued, manual stretched

would be helpful.

Hmmm, realized I have never produced these very simple and informative graphs.

These last two graphs are great Tube. The simpler they are, the more attention they will draw.

Nice work tube

Well done Tube!

The length graph really spells out for me just how far below average I started at 5” and now how I am above average at 6.8”. Man I feel good!

Tube, question about the last two graphs; how big of a sample base does that cover? It says ‘number of people’ that goes up to 40 and 80 for length and girth, respectively. Is that x10? Why the difference between the two?

It is not X10. The sample size for both graphs is 328. One of them goes up higher because their is less variance in those measurements. What I mean is that in the girth meaurements people don’t differ that much in size. Imagine if everyone measured 5 inches in girth, then the graph would have a single peak at 5 that went up to 328. The length data is “spread out” more which means each of the peaks won’t be as high.

A look at the effects of PE on our community. Recent evidence shows PE can have an effect on penis size. Look out!

Nice one, that should just about do it then.