I recently read a paper about changes in joint movement before and after therapy. They report a difference in joint movement from mean =50° to mean 0° after treatment, but a p-value of 0.05. Sample size n= 29. Maximum possible movement in that joint is 110 degrees (-20° up to 90°). In my opinion beside the difference in mean (50°) there must be many examples with less difference, otherwise the p-Value would be smaller. Any ways to conclude from the given values to the possible distribution ? The samples must have a skewed distribution, otherwise the great p-value cannot be explained. Am I right? Any ways to simulate this with R? When I use standard- normal distributed values I always end up with p-Values beneath 0.0000... Thanks Peter
5
-
1$\begingroup$ What hypothesis test did they use? $\endgroup$– Michael MNov 2, 2017 at 13:25
-
2$\begingroup$ Any idea of the standard deviations in the control and treatment groups? The difference in means is large but of the variance is also large, that would explain the p-value. $\endgroup$– KirkD_CONov 2, 2017 at 13:26
-
$\begingroup$ If they have given a confidence interval that would also help. Editors of health journals started asking for them about three decades ago but perhaps the message is yet to get through. $\endgroup$– mdeweyNov 2, 2017 at 14:37
-
$\begingroup$ what is great p-value ? you need a little more reading of literature on p-values. It does not reflect what you have in mind ? I think we have a calculated p value and a simulated p-value. $\endgroup$– user10619Nov 7, 2017 at 12:05
-
$\begingroup$ Thanks for all comments. In the paper there is no standard deviation and no CI. Very strange. Only mean and p-value. Because they only have 16 observations, I tried many combinations of values with given mean and range and calculated the respective p-values. Never ! could get a p-value greater than 0.00001. Very strange data. I am a little bit concerned about the fact, that they publish such data. $\endgroup$– Peter HahnNov 8, 2017 at 18:05
Add a comment
|