Currently, I'm working on a meta-analysis that is basically based on data obtained from individual-studies. However, there are some cases where the authors report effect size value with exactly zero standard error. I understand standard error cannot be zero & the problem is related to rounding decimal places. Is there some way to handle such cases? Any help is much appreciated!
1
-
2$\begingroup$ I don't know a way to overcome the authors' failure to use a sufficient number of significant digits in their reporting. But meta-analysis on summary data is disappointing and it's far better to acquire the raw data to do a more informative analysis that doesn't suffer from reporting format. $\endgroup$– Frank HarrellFeb 6, 2021 at 12:22
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
It might be possible to obtain the precise value of standard error from other reported characteristics. For example, if the effect sizes are computed as standardized measures, such as Cohen's d or correlation coefficient, the sample size and effect size are all the information needed for obtaining the standard error. See this question for obtaining the variance of Cohen's d - its square root is equal to the standard error. Or, if the effect size is a mean difference and the authors report the t-statistic or the p-value, you divide the effect size by the t-statistic and obtain the standard error (t = eff/se). Another option might be calculating it from sample sizes and standard deviations if there were reported and a non-standardized effect size measure was used.
If neither of those options is possible, I guess that you might try using a standard error estimate from a study that is the most similar in terms of sample size and methodology. Then check how the inclusion of this study with the "guessed" standard error affects the overall estimate and how sensitive the overall estimate is to different values of the "guessed" standard error.
-
1$\begingroup$ Thank you very much for the tips. My case is the last one. They simply put the effect size with zero standard error only. No other information is provided! $\endgroup$– iGadaFeb 6, 2021 at 16:01
-
$\begingroup$ Sorry to hear that. An incomplete reporting of results is not great. If the article is recent, you can also try contacting the authors. There is a high chance that they won't respond but it's always worth a shot. $\endgroup$ Feb 6, 2021 at 19:23