Repeated Measures ANCOVA in JASP

isakohrlund

Hi,

I want to run a repeated measures ANCOVA in JASP, but I am unsure of whether I have structured my data and entered it into the analysis in the correct way since JASP gives no guidance on this.

Specifically, what I am trying to do is to compare the mean of "X" (electricity use) between two years, while controlling for "Y" (temperature) which differs between the two years. There are 240 subjects, and each subject has 12 measurements (one per month) for each of the two years. I.e. 2880 data points are used to calculate the mean for each year. At the moment, I am only interested in the differences between years, not between months or subjects. I have structured my data as follows: Subject_ID; Measure_Year 1; Measure_Year 2; Covariate_Year 1; Covariate_Year 2

Here's a screenshot of the data structure:

Here's a screenshot of how I entered the data into the analysis and the results:

Is this the correct way to structure the data and enter it into the analysis when I have covariates? If YES, how do I interpret the results?

Thanks in advance.

Best regards
Isak

EJ

Hi Isak,
JASP assumes that every subject/unit is a row. So it assumes a "wide" format, not a "long" format. I've suggested on GitHub that we increase the priority level of including support for the "long" format. The way you've coded it, JASP is under the misapprehension that row's 1 and 2 are not from the same unit.

Also, as a side note, ANCOVA is used when a covariate adds noise but does not differ between the conditions (Miller & Chapman, 2001).

Cheers,
E.J.

isakohrlund

Thank you. Do you know of any analysis (mean comparison) that can account for "covariates" that differ between conditions?

EJ

Hi isakohrlund,

From the perspective of Miller & Chapman, the problem cannot be overcome -- it is an inherent confound. I believe Pearl had a different opinion, but there's no easy fix. So at the very least it's prudent to issue a caveat emptor.

Cheers,
E.J.

isakohrlund

Alright. I'll have to find another smart solution to this. Thank you E.J.