fix paired tests (#52)

computational-metabolomics · Jun 25, 2021 · dcb112a · dcb112a
1 parent 489170c
commit dcb112a
Show file tree

Hide file tree

Showing 4 changed files with 21 additions and 45 deletions.
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -1,7 +1,7 @@
 Package: structToolbox
 Type: Package
 Title: Data processing & analysis tools for Metabolomics and other omics
-Version: 1.4.0
+Version: 1.4.1
 Authors@R: c(
     person(
         c("Gavin","Rhys"),

diff --git a/NEWS b/NEWS
@@ -1,3 +1,6 @@
+Changes 1.4.1
++ fix broken ttest and wilcoxon test when paired = TRUE
+
 Changes 1.4.0
 + added VIP summary chart
 + add ellipse plotting options to pca_scores_plot

diff --git a/R/ttest_class.R b/R/ttest_class.R
@@ -149,26 +149,14 @@ setMethod(f="model_apply",
                 }
 
             }
-            D$data=D$data[!(D$sample_meta[[M$paired_factor]] %in% out),]
-            D$sample_meta=D$sample_meta[!(D$sample_meta[[M$paired_factor]] %in% out),]
+            #D$data=D$data[!(D$sample_meta[[M$paired_factor]] %in% out),]
+            #D$sample_meta=D$sample_meta[!(D$sample_meta[[M$paired_factor]] %in% out),]
+            D=D[!(D$sample_meta[[M$paired_factor]] %in% out),]
             y=D$sample_meta[[M$factor_names]]
 
             # sort the data by sample id so that theyre in the right order for paired ttest
-            X=apply(D$data,2,function(x) {
-                a=x[y==L[1]]
-                b=x[y==L[2]]
-                ay=y[y==L[1]]
-                by=y[y==L[2]]
-                a=a[order(ay)]
-                b=b[order(by)]
-                return(c(a,b))
-            })
-
-            # put back into DatasetExperiment object
-            D$data=as.data.frame(X)
-            D$sample_meta[[M$factor_names]]=D$sample_meta[[M$factor_names]][order(D$sample_meta[[M$paired_factor]])]
-            D$sample_meta[[M$paired_factor]]=D$sample_meta[[M$paired_factor]][order(D$sample_meta[[M$paired_factor]])]
-            y=D$sample_meta[[M$factor_names]]
+            temp=D$sample_meta[order(D$sample_meta[[M$factor_names]],D$sample_meta[[M$paired_factor]]),]
+            D=D[rownames(temp),]
 
             # check number per class
             # if less then 2 then remove
@@ -178,8 +166,7 @@ setMethod(f="model_apply",
 
             # check equal numbers per class. if not equal then exclude.
             IN=rownames(FF$count)[(FF$count[,1]==FF$count[,2]) & (FF$count[,1]>2) & (FF$count[,2]>2)]
-            D$data=D$data[,IN]
-            D$variable_meta=D$variable_meta[IN,]
+            D=D[,IN]
 
             estimate_name='estimate.mean of the differences'
         }

diff --git a/R/wilcox_test_class.R b/R/wilcox_test_class.R
@@ -121,42 +121,28 @@ setMethod(f="model_apply",
                 if (ng != 2) {
                     out=c(out,k)
                 }
-
+                
             }
-            D$data=D$data[!(D$sample_meta[[M$paired_factor]] %in% out),]
-            D$sample_meta=D$sample_meta[!(D$sample_meta[[M$paired_factor]] %in% out),]
+            #D$data=D$data[!(D$sample_meta[[M$paired_factor]] %in% out),]
+            #D$sample_meta=D$sample_meta[!(D$sample_meta[[M$paired_factor]] %in% out),]
+            D=D[!(D$sample_meta[[M$paired_factor]] %in% out),]
             y=D$sample_meta[[M$factor_names]]
-
+            
             # sort the data by sample id so that theyre in the right order for paired ttest
-            X=apply(D$data,2,function(x) {
-                a=x[y==L[1]]
-                b=x[y==L[2]]
-                ay=y[y==L[1]]
-                by=y[y==L[2]]
-                a=a[order(ay)]
-                b=b[order(by)]
-                return(c(a,b))
-            })
-
-            # put back into DatasetExperiment object
-            D$data=as.data.frame(X)
-            D$sample_meta[[M$factor_names]]=D$sample_meta[[M$factor_names]][order(D$sample_meta[[M$paired_factor]])]
-            D$sample_meta[[M$paired_factor]]=D$sample_meta[[M$paired_factor]][order(D$sample_meta[[M$paired_factor]])]
-            y=D$sample_meta[[M$factor_names]]
-
+            temp=D$sample_meta[order(D$sample_meta[[M$factor_names]],D$sample_meta[[M$paired_factor]]),]
+            D=D[rownames(temp),]
+
             # check number per class
             # if less then 2 then remove
             FF=filter_na_count(threshold=2,factor_name=M$factor_names)
             FF=model_apply(FF,D)
             D=predicted(FF)
-
+            
             # check equal numbers per class. if not equal then exclude.
-            out=FF$count[,1]!=FF$count[,2]
-            D$data=D$data[,!out]
+            IN=rownames(FF$count)[(FF$count[,1]==FF$count[,2]) & (FF$count[,1]>2) & (FF$count[,2]>2)]
+            D=D[,IN]
         }
 
-
-
         X=D$data
         y=D$sample_meta[[M$factor_names]]