Hi there,
I'm working on a systematic review but am not very proficient with stats - I have the baseline and final mean results for my control and intervention group as well as their standard deviations. However, the means differed at baseline (x - 26, y - 29), and so although the mean change from baseline is greater in one group (x - 6.6, y - 7.8), if you judge the results purely by the final mean values, the other group appears lower (x - 19, y - 21).
I would therefore like to use the 'change from baseline'parameters to calculate my measure of effect, however no change from baseline standard deviation has been provided by the authors.
I've read that I can use a correlation coefficient to calculate a SD, however the few other papers I am comparing against in the same group do not list their change from baseline SD either so I cannot use them to calculate a correlation coefficient!
Cochrane suggests using 'reasoned arguement' to come up with a correlation coefficient in this case, but I've no idea where to begin reasoning.
Any help would be much appreciated - many thanks in advance!
Two options.
1. There may be other information that will help you to derive the correlation, like in this case the t value.
2. Write to the authors.
Where is your study data and possible authors help you in this scientific communication
You can obtain a rough estimate of correlation from the other studies that reported both baseline SD, final endpoint SD, and SD of change. You can use the conversion formula forward and backward i.e
SD^2 "final endpoint" = SD^2 "change" - SD^2 "baseline + (2* coefficient r * SD "change" * SD "baseline")
Otherwise, you can assume a rough estimate of "r" between 0.6 - 0.8 and state this in your methodology. I suggest you read Chapter 16.1.3.2 in Cochrane Handbook regarding SD imputation.
https://handbook-5-1.cochrane.org/chapter_16/16_1_3_2_imputing_standard_deviations_for_changes_from_baseline.htm
In your scenario, regression to mean is the best and most accurate method to calculate the effect size. The equations provided by Mohamed Abdel-Maboud are advisable.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics