Hi, I've run several single-trait analyses using the function MPV() from the homonimous package and I'm starting to analyse the results.
I'm a bit confused by what's in the main csv output ("*.FarmCPU.csv"): according to the manual of the rMPV R package, I should get four columns: SNP_ID, Chr, Pos, and p-value.
In my case the columns are eight: SNP_ID, Chr, Pos, (as expected), REF, ALT, Effect (unexpected but I get them), SE, and phenotype_residuals.FarmCPU.
My guess is that the "phenotype_residuals.FarmCPU" column are the residuals for each model FarmCPU has ran, and "SE" is the standard error - having an option to bootstrap the pQTN for each iteration, it makes sense that the algorithm returns this statistics.
However, I expected a p-value. I used a Bonferroni correction for the field "p.threshold" by setting NA, but I gathered that it should have reported the p-value nonetheless. Furthermore, I'd be more prone to set a very lax p-value threshold and then assess the resulting p-values under a stringent FDR: the absence of the p-value column leave me perplexed.
I hypotesysed that the significantly associated" SNPs are reported in the "*_signals.FarmCPU.csv" output file, a waaay shorter file than the other. Yet, it's structurally identical to the other and still lacks the p-value column.
Someone have a clue on where to look at? Thanks in advance.
I noticed that phenotype_residuals.FarmCPU is actually FarmCPU and has nothing to do with the phenotype ~ genotype regression: i fact my phenotypes are already residual themselves. The column name should be "phenotype_FarmCPU", but, being my phenotype called thing_residual, the latter word popped up in the output. I have stronger clues that this should be the p-value column, but I'd like some opinion.