Do you have the range or medians and IQR? if so, you can estimate the SD as described in the paper  by Wan et al 2014 ("Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range").

Best,

Ahmad.

Hi Kaloyan,

I like the book "Practical meta-analysis" from Lipsey and Wilson. They describe the following possible ways to compute d:

- direct calculation based on means and standard deviations

– algebraically equivalent formulas (t-test)

– exact probability value for a t-test

– approximations based on continuous data (correlation coefficient)

– estimates of the mean difference (adjusted means, regression B

weight, gain score means)

– estimates of the pooled standard deviation (gain score standard

deviation, one-way ANOVA with 3 or more groups, ANCOVA)

– approximations based on dichotomous data

Formulas are also given. Some softwares (e.g. CMA) autamatically compute d or g through other parameters.

If you do not have sufficient information, try to contact the authors.

Best wishes and good luck,

Mira

In case of incomplete/missing data, contact the original investigators to request incomplete/missing data.

Good Luck

http://handbook.cochrane.org/chapter_16/16_1_2_general_principles_for_dealing_with_missing_data.htm

Thanks a lot, guys! I am actually using CMA, but there is no option based on the info i have. I will contact the authors, although the paper is from 1979 and i doubt the authors still keep the data. But thanks anyway and cheers!

Dear Kaloyan,

Between data extraction and calculating d or g, there is an intermediate step, which is to calculate the effect size for each study. Assuming you are dealing with continuous numeric data, there are three main ways of doing so: regular, standardized and normalized mean difference.

In any case, all you need is to have sample size, average and SD for each group. As you said, you have no SD, no p and no t. Other options would be look for some SE or confidence intervals. If you have some of this info, all calculation you may need can be found on Cochrane Handbook, Part 2, section 7.7.3.3 (Obtaining standard deviations from standard errors, confidence intervals, t values and P values for differences in means).

If this is not possible, best option would be to contact authors, or to estimate results from graphics (if available).

Last option would be to impute data (as suggested in Cochrane Handbook, Part 2, section 7.7.3.7. In this case, all information you may need can be found in the same Cochrane Handbook, part 3, section 16.1.3.1.

Bear in mind that imputations are risky, and the more you do that, the less precise your results become. So, a final bu unconvinient option would be to exclude this results from you analysis.

Good luck!

Thank you so much Gabriel!