Since in some studies, authors only report the mean difference and mean difference standard deviation. Kindly, help me out to get the absolute mean and SD of post-test measures.
Thanks
I'm not sure you can do this without the pre-test mean and the pre-test SD also. Even the test statistics wouldn't help here as they are consistent with many pre-test and post-post combinations.
Thanks Mr.Thom Baguley for your response.
I am using Revman for making forest plot, for it I require post test mean sd. I have included 11 studies, among them 8 have reported post test data in mean sd and 3 have reported in mean difference sd.
I am not getting how to formulate mean difference sd to mean sd of post test variables. So, I can make a accurate forest plot. Please suggest me, how to solve this problem.
Regards
Saima Zaki
I think the only solution is probably to contact the original study authors.
Yes, Thom Baguley it's true!
But the problem is few of author didn't revert back. I am waiting for their responses.
Dear Saima Zaki
We would expect some correlation between the baseline score and the final score in and intervention and therefore the variation on the change score and the SDchange may be somewhat smaller than the SDfinal In order to estimate the SDfinal we can use the formula from the
Cochrane handbook 6.5.2.8 Imputing standard deviations for changes from baseline
SD = SQRT (SDbaseline *SDbaseline + SDfinal*SDfinal – 2*corr* SDbaseline* SDfinal)
– this leaves you with another challenge – what is the correlation? Corr is the correlation between the baseline score and final score and this may depend on the intervention and especially the time between baseline and the time to estimate the final score. However, as all the studies in your meta-analysis has the same intervention and time (or at least so close that you would include them in the same meta-analyses) – you can therefore use the correlation from other studies in your meta-analysis (and if it can be estimated from more than one study then use the median of the correlations estimated from the other studies in the meta-analyses)
Best Carsten
Respected Carsten Juhl, thank you so much for your kind gesture.
I have furthermore queries, I have read some of the Cochrane meta-analysis. In the Forest plot, I have observed they have reported mean SD or mean SE or mean difference SD. Then by the Revman final standardized mean difference SD and CI mentioned in Forest plot.
For my meta-analysis, I have planned to use mean SD of post-test data of my eligible studies for making the Forest plot. For most of my included eligible studies, I got the post-test mean SD, but for those studies in which post-data is reported in other than mean SD (i.e., mean differences or median IQR), I am having problems.
Although, I have found calculator to derive statistical data in Revman. But in my further analyzing I found, this calculator is not valid (not giving appropriate values).
In summary, I just want a valid equation to get post-test mean SD from mean difference SD and median change IQR. I have baseline data too.
Thanks
Warm Regards
Saima Zaki
Dear Saima Zeki
If you have the mean difference (MD) and the standard difference on the mean difference (SD) you can look at chapter in the Cochrane handbook
16.4.6.1 Mean differences
The standard error of the mean difference is obtained as
SE (MD) = SDdiff/sqrt(N)
where N is the number of participants in the trial, and SDdiff is the standard deviation of within-participant differences between Experimental and Control measurements.
Median and interquartile are discussed in
7.7.3.5 Medians and interquartile ranges stating
When sample sizes are large and the distribution of the outcome is similar to the normal distribution, the width of the interquartile range will be approximately 1.35 standard deviations.
Best Carsten
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis