Is there a way to do meta-analysis for crossed designs (no control group)? I have a 2x2 factorial with count data that I would be interested in looking at. I basically did the same experiment five times and would like to look at the results across all five experiments simultaneously. Meta-analysis strikes me as the best option, but I've only done it to compare two groups before.

Further details on the study design if you are interested:

Factor 1: Soil type (two levels, organic or conventional)

Factor 2: Seed type (two levels, transgenic or not)

Soil used was from 5 different pairs of soil samples from organic and conventional fields and the experiment was run once for each pair. Or, to clarify:

Exp 1: Field 1 vs. Field 2

Exp 2: Field 3 vs. Field 4

Exp 3: Field 5 vs. Field 6

Exp 4: Field 7 vs. Field 8

Exp 5: Field 9 vs. Field 10

Where the odd numbered field is conventional and the even numbered field is organic each time.

Is meta-analysis the best way to combine results from these experiments? Is there another way? If meta-analysis is the best way, how do I go about doing it with four groups and no control?

  • $\begingroup$ Since you have the raw data why not analyse that directly rather than using a two-step approach of calculating a summary statistic and meta-analysing it? $\endgroup$ – mdewey Jul 16 '16 at 13:47
  • $\begingroup$ Since each experiment was run separately, I have been discouraged from doing it this way. Could I potentially combine them using both the rep within an experiment and the experiment itself as random variables? $\endgroup$ – user3391456 Jul 22 '16 at 18:36
  • $\begingroup$ Centres in a multi-centre trial are run separately (with separate randomisation lists) but analysed together. $\endgroup$ – mdewey Jul 22 '16 at 20:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.