I am new to cyclic data analysis, and some collaborators suggested the use of JTK to test for rhythmicity in data.
I have metabolite quantities in mice feces across 20h (sampled every 4h), and I want to know which are cycling and which do not. I chose to test this with the MetaCycle package in R, because the metad3d function, which takes individuals data and integrate them in a global test, seemed better than looking at means across individuals. I tried running it with JTK and ARS, but I get opposite results: with JTK nothing reaches significance (and it's really far, most p-values are above 0.8), and in ARS pretty much everything is significant (and most p-values are below 0.01). I know that different statistics are bound to yield different results, but they should still be pointing in the same direction.
Does anybody has an idea why I get such opposite results ?
There is potential explanation:
For JTK, 4h sampling is a minimum and 2 cycles are required. Not well adapted with only 1 cycle. ARSER did not consider replicates and can't cope with missing data that are likely present with metabolomic data. How many replicates do you have? Is the sampling in DD or LD conditions?
For sure, never work with means! If you have 4 replicates, you can consider you have 4 cycles with 1 replicate. It will increase the power of your analysis.
- Badges
- Science topic