Dear fellow Stata users,
thank you very much for reading my post. I am currently working on an interesting analysis and asking for your advice. From what I read, an approach including bootstrapping / Monte Carlo simulation might help me, but, unfortunately, I seem to be unable to find the right source to help guide my work.
I have conducted an experiment with three groups:
In group 1, participants had to answer a task (baseline task).
In group 2, participants received some sort of intervention and then had to answer a slightly modified task (treatment task).
In group 3, participants first had to answer the baseline task, then received the above-mentioned intervention, and were allowed to change their previous answer (comparable to the treatment task).
I have an ordinal variable from each task (one for each participant group 1, one for each participant group 2, two for each participant in group 3) as well as an (almost) numerical equivalent of this ordinal variable (based on specific utility function, some participants are excluded due to this transformation if their choice was irrational). I use this numerical variable for my pairwise testing of the treatment effects. For that, I used between-subject nonparametric testing (Mann-Whitney-U-tests between group 1 and group 2 as well as group 1 and the second answer in group 3). In addition, I performed within-subject nonparametric testing (Wilcoxon-Signed-Ranks-test between the first and the second answer in group 3). With that, I can investigate the treatment effect (change between group 2 and group 1) as well as the treatment effect after an initial baseline decision (change between the second and the first answer in group 3).
I want to follow up on this with a nonparametric test between the "difference between group 1 and group 2" (baseline-treatment) and the "difference within group 3" (first answer-second answer). I thought about randomly matching participants from group 1 with participants in group 2, calculating the pair's difference, using these differences for a comparison with the differences in group 3.
Does this approach sound like a reasonable idea to you?
Do you have any source / paper / advice / guidance on how to perform this in Stata?
thank you very much for reading my post. I am currently working on an interesting analysis and asking for your advice. From what I read, an approach including bootstrapping / Monte Carlo simulation might help me, but, unfortunately, I seem to be unable to find the right source to help guide my work.
I have conducted an experiment with three groups:
In group 1, participants had to answer a task (baseline task).
In group 2, participants received some sort of intervention and then had to answer a slightly modified task (treatment task).
In group 3, participants first had to answer the baseline task, then received the above-mentioned intervention, and were allowed to change their previous answer (comparable to the treatment task).
I have an ordinal variable from each task (one for each participant group 1, one for each participant group 2, two for each participant in group 3) as well as an (almost) numerical equivalent of this ordinal variable (based on specific utility function, some participants are excluded due to this transformation if their choice was irrational). I use this numerical variable for my pairwise testing of the treatment effects. For that, I used between-subject nonparametric testing (Mann-Whitney-U-tests between group 1 and group 2 as well as group 1 and the second answer in group 3). In addition, I performed within-subject nonparametric testing (Wilcoxon-Signed-Ranks-test between the first and the second answer in group 3). With that, I can investigate the treatment effect (change between group 2 and group 1) as well as the treatment effect after an initial baseline decision (change between the second and the first answer in group 3).
I want to follow up on this with a nonparametric test between the "difference between group 1 and group 2" (baseline-treatment) and the "difference within group 3" (first answer-second answer). I thought about randomly matching participants from group 1 with participants in group 2, calculating the pair's difference, using these differences for a comparison with the differences in group 3.
Does this approach sound like a reasonable idea to you?
Do you have any source / paper / advice / guidance on how to perform this in Stata?
Comment