Linear Mixed Effects on the First differences interpretation

Dear all,
I always had problems with LME output; it makes me cry :sob:. By doing it with first differences, I am completely lost.
So, I have data with 3 time points: 0 (baseline tp), 2 and 12 (intervention), and 3 diets: diet_A, diet_B, and baseline_C (baseline diet). First, I calculated first differences by providing tp 0 as a baseline. Then I used this output for LME.
Here is LME output:
0-12W-shannon-first-differences-lme.qzv (312.6 KB)

Am I correct that in diet_B Shannon index differences to tp 0 were significantly different from differences to tp 0 in baseline_C overtime with negative correlation? So, interpretation will be that while shannon index increased in baseline at tp 2, it decreased to tp 12, and in diet_B it decreased to tp 2 and increased to tp 12? :crazy_face:

2 Likes

Hi @timanix ,
Sounds like you are not lost at all :grin:

The term LME_Group[T.diet_B] without time interaction is indicating a significant difference from the control (baseline_C) at the intercept. As you are running this on shannon first differences, you are correct that this would indicate that Shannon increased in the baseline_C group between week 2 and week 0, but decreased in the diet_B group.

No this part does not sound right. First differences is looking at the rate of change, and so it looks like the average rates of change are stable in all groups across time (i.e., constant positive change in baseline_C and constant negative in diet_B, no change in diet_A), but you would need to look at the Shannon diversity volatility directly to confirm that (i.e., not first differences).

3 Likes

Thank you for the explanation! Now it is clear, I will check the volatility.

1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.