search

strengejacke / esc

Effect Size Computation for Meta Analysis
https://strengejacke.github.io/esc
20 stars 6 forks source link

Is the formula used for esc package and comprehensive meta-analysis software different? #16

Open Elie-Yu opened 3 months ago

Elie-Yu commented 3 months ago

I noticed there are different results when calculating effect size with information available for both treatment and control groups on sample size and the mean and SD for pretest and posttest. this comparison is between using the esc package and comprehensive meta-analysis software.

An example of data and script that I used is

In esc package: esc_mean_gain(pre1mean = 3.34, pre1sd = 1.76, post1mean = 3.28, post1sd = 2.07, grp1n = 34, pre2mean = 3.81, pre2sd = 1.79, post2mean = 8.43, post2sd = 0.83, grp2n = 34, es.type = "g") The results returns g= 2.755 var(g) = 0.3114

Using the same values in Comprehensive Meta-analysis, the results returned as follows: g= 2.9338 var(g) = 0.343

I wonder if different formula is used. Could you help me understand what leads to the difference and what results to choose from?

Elie-Yu commented 3 months ago

To follow up, I also did the calculation in "Practical Meta-Analysis Effect Size Calculator"

using different SDs:

Screenshot 2024-05-18 at 11 46 00 AM

<img width="564" alt="Screenshot 2024-05-18 at 11 46 11 AM" src="https://github.com/strengejacke/esc/ass

Screenshot 2024-05-18 at 11 46 34 AM

ets/140570906/3d218406-8e25-4531-b04d-8424795b160f">