Doesn't seem possible to me.
Hi there,
I am trying to calculate cohen's d (standardised mean difference) for a study that only reports the range (min and max), and not the mean or standard deviation. I am wondering if it's possible to calculate cohen's d using the sample size and range? And if so, what formula will provide the best estimate of d?
Thanks so much,
Dee
Doesn't seem possible to me.
Stop cowardice, ban guns!
Very roughly sd = (max - min)/4 and mean = (max + min)/2. This may be enough for a rough estimate of Cohen's d, as effect size is a fairly vague measure for a start.
Perhaps this paper could help. Estimating the mean and variance from the median, range, and the size of a sample. Hozo et al
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1097734/
Nice katxt. Haven't read the link yet but I would imagine some assumptions are needed. I wonder if a bunch of simulations could be of use here as well.
I would recommend the standard meta-analysis sensitivity analysis as well, where you run the MA with leave one study out until you have gone through the permutation of included studies.
Stop cowardice, ban guns!
Thanks so much for your feedback - the reference is incredibly helpful. I will make note to run the sensitivity analysis too.
Thanks again,
d
Tweet |