Noi/NOTES/2012-6-14

From ZhangLabWiki
Jump to navigation Jump to search

Regression analysis of UCLA SZ data (disease status)[edit]

  • Continued from: http://genome-tech.ucsd.edu/LabNotes/index.php/Noi/NOTES/2012-6-13
  • From the comment on previous page
    • LDA: random choose samples for training and (have UCLA-SZ and UCSD-GA on training and testing data sets)
  • Training and testing data set (I used the same group of training and testing data as assigned for age analysis)
    • list 1-96 =>UCSD-SZ, 97-169 => UCSD-GA
    • trainAge <- c(1:40, 91:132)
    • testAge <- c(41:90, 133:169)
  • From the previous analysis, I used PC1-PC3 scores as input for LDA analysis. However when I plot scree plot of the variances contributed on each PC, the change in slope in the scree plot occurs at component 4. In this analysis, I'll put PC1-PC4 scores as input.
A=read.table("UCLA_regression_status_pval0.015_MM",header=TRUE,row.names=1)
B=na.omit(A)
B$STDEV=NULL
B$min_RD=NULL
B$mean_RD=NULL
pca1 <- prcomp(t(B), scale=TRUE)
screeplot(pca1, type="line") 
File:Screeplot status pval0.015.png
Script: lda_status_PC1-4_v2.R
File:Plot LD1 status PC1-4 p.015 v2 training.png File:Plot LD1 status PC1-4 p.015 v2 test.png

Table of raw data

Status (p-value 0.015) Predicted\Truth Affected Unaffected Correct FALSE Total
Training Affected 10 0 82 0 82
Unaffected 0 72
Testing Affected 10 2 83 4 87
Unaffected 2 73
Cross validation Affected 19 2 164 5 169
Unaffected 3 145

Table of accuracy and misclassification rate

Training (Accuracy of allocation) Misclassicication rate Testing (quality of model) Misclassicication rate Cross validation Misclassicication rate
82 of 82 0.00% 83 of 87 4.60% 164 of 169 2.96%

Compare the result to previous analysis by treating UCLA-SZ as training data and UCSD-GA as testing data

Training (Accuracy of allocation) Misclassicication rate Testing (quality of model) Misclassicication rate Cross validation Misclassicication rate
92 of 96 4.17% 73 of 73 0.00% 164 of 169 2.96%

Summary