Warnings in CpGassoc "sites were not able to be analyzed"

60 views
Skip to first unread message

izab...@sas.upenn.edu

unread,
Sep 5, 2018, 12:10:08 PM9/5/18
to Epigenomics forum
I used the following code to try and run an EWAS:

ewas <- cpg.assoc (betas, phenotype_final$independent_var, covariates)

And keep getting these error messages where it says all of the CpG sites were not able to be analyzed. 


Warning messages:
1: In rm(non.m.beta, sser, ssef, beta0, r.ressq) : object 'sser' not found
2: In rm(non.m.beta, sser, ssef, beta0, r.ressq) : object 'ssef' not found
3: In rm(non.m.beta, sser, ssef, beta0, r.ressq) :
  object 'beta0' not found
4: In rm(non.m.beta, sser, ssef, beta0, r.ressq) :
  object 'r.ressq' not found
5: In cpg.work(beta.val, indep, covariates, data, logit.transform,  :
  7207 sites were not able to be analysed.
Test statistics for these were set to NA
6: In min(test.stat$P.value, na.rm = TRUE) :
  no non-missing arguments to min; returning Inf


From the source code of CpG assoc for #5, where levin is the independent variable as a factor: 

  if(!levin) {
    test.stat[nonmissing,1]<-sqrt(test.stat[nonmissing,1])*sign(beta[2,])
    if(beta.col>1) { 
    e.s[nonmissing,]<-cbind(colSums(colMeans(r.design)*matrix(beta[-2,],ncol(r.design))),beta[2,])
       }
    else {
      e.s[nonmissing,]<-cbind(colSums(as.matrix(colMeans(r.design)*matrix(beta[-2,],ncol(r.design)))),beta[2,])
      }
    stderror[nonmissing,]<-abs(e.s[nonmissing,2]/test.stat[nonmissing,1])
       }}
   if(callarge) {
   rm(beta)
   gc()}
   if(sum(is.na(test.stat[,1]))>0) {
      warning(sum(is.na(test.stat[,1]))," sites were not able to be analysed.\n",
          "Test statistics for these were set to NA")
      }  
     
     }

Can someone explain the source code to me and what the problem is?
Reply all
Reply to author
Forward
0 new messages