0
$\begingroup$

I'm trying to analyze data from an intervention study that has 70 participants separated into two groups (Control and Intervention - Between Subjects) and looks at pre-post data (Within Subjects) The dependent variable is the number of times a certain document was accessed. Prior to the intervention, baseline data was examined in both groups, the intervention was implemented, and then the number of times the document was accessed was examined again to see whether there were difference between the control and intervention group.

        BASELINE | POST 
CONTROL     6       17
INTERV      3       11

Considering we have 1 between variables (group) and 1 within variable (time), with frequency/count data as the DV, what's the best analysis for this time of data? I know we can't do Chi-Squares because of the within-subjects metric.

Edit: To answer some questions below, I have individual subject data (i.e. if the individual accessed the data at baseline, and if they accessed it post-int). The group were randomized; n = 35 for the control and n = 36 for the intervention group. I can't explain why Control accessed the doc more during baseline, other than simple variance. Any help would be appreciated, thanks.

Improve this question
$\endgroup$
2
  • $\begingroup$ How do you explain that Control subjects accessed the document more often at baseline and afterwards? Were the 70 subjects split 35:35? $\endgroup$
    – BruceET
    Commented Nov 26, 2019 at 21:08
  • $\begingroup$ Do you have data on each subject or only on the groups? $\endgroup$
    – Peter Flom
    Commented Nov 27, 2019 at 11:24

1 Answer 1

Reset to default
0
$\begingroup$

I suggest a nonlinear multilevel model with a logistic link.

One level is the individual and the other level is the group. Although it's better to have more time points, these models sometimes work well even with only two time points.

Improve this answer
$\endgroup$
2
  • $\begingroup$ That does seem like the path forward - I was wondering if you knew which package would be best beneficial for this in R. I can find lots on non-linear multilevel (or mixed) models, but am having trouble finding incorporating the logistic link. $\endgroup$
    – Dhawal Selarka
    Commented Dec 2, 2019 at 14:47
  • $\begingroup$ I mostly use SAS for stuff like this, but I think the nlme package is what you want. $\endgroup$
    – Peter Flom
    Commented Dec 2, 2019 at 16:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.