I think I understand now that you ignore the baseline (pre-test) data when calculating Hedge's g. Instead subtracting the intervention posttreatment mean from the control posttreatment mean and diving the result by the pooled standard deviation (of both samples at posttreatment). However, I am now wondering that means for studies where, despite randomisation, there were significant differences in the outcome of interest at baseline.
For example, with the data below, where let's say the mean value (M) is pertaining to a depression score. Would it be appropriate to calculate Hedge's g in the method above or what would have to be done differently, if intervention and control group baselines scores were not similar?
Thankfully I think only a couple studies had this problem, but I am unsure whether I exclude, perform a correction, or run as normal in the meta-analysis.
Intervention Group Pre-treatment: M=63.92; SD=10.67; N=63
Intervention Group Post-treatment: M=59.43; SD=7.23; N=63
Control Group Pre-treatment: M=74.57; SD=9.79; N=65
Control Group Post-treatment: M=72.69; SD=4.84; N=65
Many thanks for any help.
Hello John,
Ideally, an ancova design would be applied (with pretest scores as the covariate), and adjusted/estimated posttest mean scores would then be used to compare the two groups. If that's not what the researchers did, and you don't have access to the data (if not the raw data then minimally, you'd need the pre-post correlation for each group and overall in addition to what you have), then there's no way to apply that yourself.
Good luck with your work.
Dear John,
Please, read my paper and see some options:
Article The Effects of Resistance Training on Architecture and Volum...
Regards,
Gokhan
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression