codeOriginalDraft

canData <- read.csv(file="/Users/carleebettler/Downloads/breastCancerData.csv", header=TRUE, sep=",")

canData$x <- NULL

canData$id <- NULL

canData <- canData[,1:31]

#boxplot(canData$radius_mean)
        
#boxplot(canData$area_mean)

canData$diagnosis <- ifelse(canData$diagnosis == "M",1, 0)

logit <- function(p) log(p/(1-p))

modelCan <- glm(diagnosis ~ .,family = binomial(logit), data=canData)

summary(modelCan)

pr <- predict(modelCan, canData, type = "response")

table(actual = canData$diagnosis, predicted = pr > .5 )

#---------------------------- DFA:
library(MASS)

fit <- lda(diagnosis ~ ., data=canData, na.action="na.omit", CV=TRUE)

ct <- table(canData$diagnosis, fit$class)
diag(prop.table(ct, 1))

sum(diag(prop.table(ct)))

#--------------------------- Random forest

#Paul add here 
#--------------------------- PCA
library(ggbiplot)

dataRefined <-log(canData[,2:31])

dataRefined <- do.call(data.frame, lapply(dataRefined, function(x) {
  replace(x, is.infinite(x), 9999999)
  })
)

pca <- prcomp(dataRefined, center=TRUE, scale. = TRUE)

summary(pca, loadings=T)

library(devtools)

gPlot <- ggbiplot(pca, obs.scale = 1, groups = canData$diagnosis, var.scale = 1, ellipse = TRUE, circle = TRUE)

gPlot