different statistical significance everytime odp() function is called  #22

Activity

1. ajbass commented on Oct 29, 2019

   

   ajbass

   on Oct 29, 2019

   Collaborator

   More actions

   The odp function runs a bootstrap algorithm to calculate the null distribution (to get p-values). It really depends on the number of tests (or genes) you have. Here there are not many tests, and so you would want to run a lot of bootstraps. I would recommend increasing bs.its:

   odp(de_obj, verbose=FALSE, n.mods=50, bs.its = 1000) 
   

   In the vignette we use a very small subset of the actual data to demonstrate how to use the commands.
2. sridewi1 commented on Oct 30, 2019

   

   sridewi1

   on Oct 30, 2019

   Author

   More actions

   Thank you for the prompt reply.
   I did just as you suggested and go even higher to bs.its = 1000000, but I still got two different summaries. My dataset consists of 495 features in 35 different samples. How high should or can I go with the bootstrap iteration?

   ExpressionSet Summary
   
   ExpressionSet (storageMode: lockedEnvironment)
   assayData: 495 features, 35 samples
     element names: exprs
   protocolData: none
   phenoData
     sampleNames: E14F1 E14F2 ... POM6 (35 total)
     varLabels: adj.var tme
     varMetadata: labelDescription
   featureData: none
   experimentData: use 'experimentData(object)'
   Annotation:
   
   de Analysis Summary
   
   Total number of arrays: 35
   Total number of probes: 495
   
   Biological variables:
           Null Model:~adj.var
   <environment: 0x56465c3c5748>
   
           Full Model:~adj.var + ns(tme, df = 2, intercept = FALSE)
   <environment: 0x56465c3c5748>
   
   .......
   
   
   Statistical significance summary:
   pi0:    0.1498209
   
   Cumulative number of significant calls:
   
             <1e-04 <0.001 <0.01 <0.025 <0.05 <0.1  <1
   p-value      134    200   282    311   345  375 495
   q-value      152    245   342    383   428  464 495
   local fdr     97    171   259    287   312  349 495
   

   the second result using the same parameter bs.its=1000000

   
   ExpressionSet Summary
   
   ExpressionSet (storageMode: lockedEnvironment)
   assayData: 495 features, 35 samples
     element names: exprs
   protocolData: none
   phenoData
     sampleNames: E14F1 E14F2 ... POM6 (35 total)
     varLabels: adj.var tme
     varMetadata: labelDescription
   featureData: none
   experimentData: use 'experimentData(object)'
   Annotation:
   
   de Analysis Summary
   
   Total number of arrays: 35
   Total number of probes: 495
   
   Biological variables:
           Null Model:~adj.var
   <environment: 0x56465c3c5748>
   
           Full Model:~adj.var + ns(tme, df = 2, intercept = FALSE)
   <environment: 0x56465c3c5748>
   
   .......
   
   
   Statistical significance summary:
   pi0:    0.1248155
   
   Cumulative number of significant calls:
   
             <1e-04 <0.001 <0.01 <0.025 <0.05 <0.1  <1
   p-value      133    198   282    322   345  371 495
   q-value      157    249   353    402   442  479 495
   local fdr    100    175   264    305   333  365 495