Which test please - Ordinal data, two independent groups measured at two time points.

#1
Hello,

As part of an RCT with 100 participants in each arm we use a 5-point likert scale to assess change an outcome over time in each arm (two independent groups, each measured at two time points).

We therefore have baseline and end of study data for two independent groups and would like to assess which group showed greatest improvement over time. I hear different arguments regarding if you can/cannot use a mixed-model ANOVA for likert data, but I am not sure of the non-parametric equivalent? As there are 100 responses each time, I have also seen that the mean result could start to resemble scale-type data, and therefore a parametric test is absolutely fine.

Any advice would be appreciated.
 

Karabiner

TS Contributor
#2
Do you mean that you have a single 5-point Likert item as outcome measure?
Or do you mean that you have a Likert-scale (which consists of several
Likert-items, and its sum score is often considered quasi as an interval scaled
measure)?

With kind regards

Karabiner
 

Karabiner

TS Contributor
#4
A RCT with nothing more than a single 5-point Likert item as outcome measure, this is a bit surprising.
You can neither measure very reliably, nor differentiate well with such a measure. Do you have some other
instruments for outcome measurements also?

IMHO you have to options with that item. Either to use GEE (generalized estimating equations), which can
handle repeated-measures with ordinal depenedent variable. Or to perfom an ordinal logistic regression,
including the basline measure as predictor and the follow-up measure as dependent variable.

With kind regards

Karabiner