I have RNA-Seq count data that are expressed over five different time points. However, there is no replicate.
So I used exactTest and did pairwise comparisons.
Regarding the dispersion estimation, in case of no replicate, the edgeR manual suggests to use GLMCommonDisp with options "robust=TRUE, subset=NULL, method=deviance".
After following this, when I see the estimated common dispersion, some had common dispersion 0.1~0.2, some had common dispersion 0.5~0.6, others had 0.9. (Similar results were also obtained in tagwise dispersions.)
Dispersions over 0.4 looks pretty bad. Would it be okay to proceed to analyze this data with these common dispersions? If not, which way is recommended to estimate common/tagwise dispersions?
Thank you in advance.
So I used exactTest and did pairwise comparisons.
Regarding the dispersion estimation, in case of no replicate, the edgeR manual suggests to use GLMCommonDisp with options "robust=TRUE, subset=NULL, method=deviance".
After following this, when I see the estimated common dispersion, some had common dispersion 0.1~0.2, some had common dispersion 0.5~0.6, others had 0.9. (Similar results were also obtained in tagwise dispersions.)
Dispersions over 0.4 looks pretty bad. Would it be okay to proceed to analyze this data with these common dispersions? If not, which way is recommended to estimate common/tagwise dispersions?
Thank you in advance.