Originally Posted by bohm
Okay, I hope it’s this pump study:"A vacuum device for penile elongation: fact or fiction?"
"Of the 31 men, 27 returned for follow-up at a mean (SD) of 8 (2.5) months. The mean (range) age was 24 (18–35) years. The median stretched penile length was 7.6 (6.9–9.4) cm before treatment and 7.9 (7–9.7) cm after 6 months of vacuum therapy; there was no significant difference between the length before and after treatment (P=0.2). The efficacy of the method was 11.1% (defined as an increase of ≥ 1 cm in stretched penile length)."
http://onlineli brary.wiley.com … 006.05992.x/pdf
Yes that is the study.
The so-called p value or confidence interval determines the likelihood of making a type I (or alpha) statistical error which is to reject the "null hypothesis" when it is in fact true. In this case, the null hypothesis is that true average stretched flaccid penis length will be the same as the average starting length in this group of men after 6 months of pumping for 1 hour a week.
In this particular paper, the researchers only say that the applied "student’s t-test" and that p=.20. They do not say whether they applied a one-tailed t-test or a two-tailed t-test. A two-tailed t-test would have allowed for the possibility that pumping might actually have resulted in shortening of the penis, but it would have required a larger sample size to achieve equal power for any given confidence interval. My guess is they applied a one-tailed test.
Most researchers set the bar very high for type I statistical errors. That is why p values of < .05, < .01, or even < .001 are often considered "statistically significant" to "highly statistically significant". In this case, using a p < .05 allows for a 1 in 20 chance of a Type I error, p < .01 a 1 in 100 chance of a Type I error, and p < .001 a 1 in 1000 chance of a Type I error. In other words, a Type I error would be to conclude that pumping does elongate the penis when in fact it does not.
But a p value equal to .20 would be equivalent to accepting a 1 in 5 chance of a type I error. But conversely it implies that there is still up to a 4 out of 5 chance that pumping did in fact, result in elongation.
To calculate the statistical power of this study would require not only to establish what confidence interval was appropriate and whether a one or two-tailed t-test was used, but also an estimate of standard deviation of measured flaccid stretched penis length for repeated measurements for this group of men. The study indicates that the average stretched flaccid penis length was in fact 7.9 cm at the conclusion of the study compared to 7.6 cm at the start which is about a 4 % increase. The conclusion that this difference was "not statistically significant" implies that the researchers attributed the difference as being accounted for by measuring error, or within the expected range of variability for repeated measurements. We have no way of knowing whether the researchers actually did any type of analysis on variability of repeated measurements of flaccid stretched length in this, or any other group of men. If they did not, they have no way of estimating the power of the study. The power of a study determines the probability that an incorrect null hypothesis is appropriately rejected.
What we are interested in here is the probability of a type II (beta) statistical error. That is accepting the null hypothesis when it is in fact false, or in this case, incorrectly concluding that pumping for one hour per week for 6 months does not result in an increase in stretched, flaccid penis length. If we could calculate the power for this study, the likelihood of a type II error would be known since: power = 1 - probability(Type II error).
So the researchers conclude that a pretty lame pumping regimen (1 hour total per week without any other form of PE) conducted for 6 months with an observed increase in stretched flaccid length of 4% is not an effective treatment for penile elongation because of a p=.20. My conclusion would be that a p=.20 suggests that a larger sample size in a study of longer duration with a more rigorous regimen may very well have yielded "statistically significant" results.
One can argue that if in fact a type II statistical error was made here due to limited sample size, that pumping is still "ineffective" because it resulted in only a 4% increase in penis length after 6 months. My counter argument would be that a lot of guys here would be happy to find a PE regimen that would consume only one hour a week which resulted in a 4% increase in penis length in 6 months time.