# Imputing standard deviations for changes from baseline

This is more of a further question on a previous topic, so my apologies if there was some way to link them that I haven't found. It is essentially the same problem as How to back-calculate change from baseline from a p-value for a paired t test . I.e. I need to calculate the standard deviation of the change. However in the study I have the two samples are indepedent, and therefore the authors have already calculated a p-value using the two-sided Wilcoxon rank sum test. Specifically:

Sample 1: n=86, mean=0.58, sd = +/- 0.12

Sample 2: n=69, mean = 0.41 sd = +/- 0.108

The authors have used a two-sided Wilcoxon rank sum test to get a p-value of 0.51.

Is is possible to impute the standard deviation of the change? This is necessary for a meta-analysis I am writing.

Many thanks for any pointers.

-

It is not possible without an estimate of the correlation between baseline and follow-up measurement. Note that change scores have many problems as detailed in http://biostat.mc.vanderbilt.edu/ManuscriptChecklist. In addition, in a randomized study if the correlation between baseline and follow-up measurement is less than 0.5, it is less efficient to subtract baseline than to ignore it. The same is not true for covariate adjustment.

-