How to compare different BQSR runs?
I am trying BQSR with different known SNP files. How do I know if one BQSR run is better than the other?
Is RMSE the right criteria, ie the smaller the better?
I am calculating RMSE by summing (ObservationsAccuracyAccuracy) for one category (e.g. QualityScore for Base Substitution) in the csv generated by AnalyzeCovariates, then I divide it by total Observations and square root it to obtain RMSE. Is this the right way to calculate RMSE?
I noticed that different runs gave me different number of total Observations. Why is that?
What about Errors in the csv file? What does it mean? Can it be used to determine which BQSR run is better?
Thanks a lot in advance!