FAQs on the coding template for meta-analysis


Does an orthogonal design of 2x2x2 require treatment variables or one with eight values?

If you have three binary variables in your design, you need to have three study (or sample, if for example participant sex is one factor) variables describing these factors. For each of the eight possible combinations of the variable values, we ideally have one row on the level of the outcomes.
A minimal example:

(1 - Cognitive, 2 - Behavioral)
(1- male,2 - female)
(duration therapy:
1 - 2 months,2 - 6 months)
Outcome (for example mean score of the group)
1 1 1 8,5
2 1 2 7,8
1 2 1 8,3
... ... ... ...


Yes, the sample description then has to be done for each group.

The response rate is the share of usable questionnaires (or study information on conducted experiments) from all eligible participants invited. This might not be given in each study (for example, if participants are recruited via flyers, we do not know, how many saw the flyer and there is no number of individual contact attempts). Then, the variable is not compulsory.


If your independent variables are categorical, specify separate groups differing regarding these variables and characterize each group concerning these variables, which you have to specify in the template. For each of the possible combinations of the variable values, we ideally have one row on the level of the outcomes. If you want to depict interactions with continuous independent variables, you also need to give summary information (means) for each separate group on the continuous variable. If there is information for many samples in a meta-analysis in the end, the continuous variable can be used as a moderator to explain differences in the outcomes.
A minimal example:

(0 - male, 1 - female)
(1- cognitive,2 - behavioral)
Age(in years)
Outcome (for example mean score of the group)
0 1 40,3 8,5
1 2 35,6 7,8
1 1 31,9 8,3
0 2 52,3 8,7

My experimental design is simple: 1 IV with 2 experimental groups, between participants.

Perhaps a minimal example for the outcome-sheet with one IV (let’s assume it to be binary, e.g. medication 1 and medication 2) and 2 samples (1 for each medication) might help:

report_ID study_ID sample_ID outcome_ID o_yi
(e.g. group mean on Beck Anxiety Inventory (BAI))
o_sei o_ni
1 1 1 1 30,5 5,6 43
1 2 2 2 32,2 6,3 39

Do I enter these outcome variables in the first section of that worksheet (Outcome measures for two-group comparisons) as measures of quantitative variables?
And do I delete the rest of sections/cells?

For three dependent variables, you will have three rows with three different outcome_IDs. For a two-group-comparison of continuous variables, you will need to fill in the cells shown in the example. You can delete the other cells. Of course, in this tiny example, we have only one treatment and one sample, as you can see from the IDs!

report_ID study_ID sample_ID outcome_ID o_group1 o_m1i o_sd1i o_n1i o_group2 o_m2i o_sd2i o_n2i
1 1 1 1 Control 5,6 1,5 54 Treat 3,4 1,3 53
1 1 1 2 Control 4,3 1,3 54 Treat 4,2 1,1 53
1 1 1 3 Control 2,5 2,1 53 Treat 2,7 1,5 53

There are no blank cells where we should enter our study information.

If you have further variables, that you would like to add and that are not specified in the template, you can do the following: At first, decide at which level of analysis your variable of interest belongs and go to the corresponding tab. According to the naming convention, a variable describing the treatment will be in the tab “Lev3Treatment” and will be called “t_[variable]”. Write the variable in the first empty column. To give us an understanding of what your variable measures / means, please also describe your new variable in the “Codebook”-tab. We have added a few blank rows for each analysis level, that you may make use of.