set.seed(12)
df <- read.csv("diabetes.csv")
head(df)
sample.data<-sample.int(nrow(df), floor(.50*nrow(df)), replace = F)
train<-df[sample.data, ]
test<-df[-sample.data, ]
Try to predict glucose from the other columns
## Warning: Removed 7 rows containing missing values (`geom_point()`).
## Warning: Removed 3 rows containing missing values (`geom_point()`).
## Warning: Removed 3 rows containing missing values (`geom_point()`).
Make a new column to determine diabetes or not and then do boxplots
train$diabetes<-ifelse(train$glyhb>6.5, 1, 0)
train$diabetes<-as.factor(train$diabetes)
df1<-train[!is.na(train$glyhb),]
ggplot2::ggplot(df1, aes(x=diabetes, y=chol))+geom_boxplot()
labs(x="diabetes", y="chol")
## $x
## [1] "diabetes"
##
## $y
## [1] "chol"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=stab.glu))+geom_boxplot()
labs(x="diabetes", y="stab.glu")
## $x
## [1] "diabetes"
##
## $y
## [1] "stab.glu"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=hdl))+geom_boxplot()
labs(x="diabetes", y="hdl")
## $x
## [1] "diabetes"
##
## $y
## [1] "hdl"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=ratio))+geom_boxplot()
labs(x="diabetes", y="ratio")
## $x
## [1] "diabetes"
##
## $y
## [1] "ratio"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=height))+geom_boxplot()
## Warning: Removed 2 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="height")
## $x
## [1] "diabetes"
##
## $y
## [1] "height"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=weight))+geom_boxplot()
labs(x="diabetes", y="weight")
## $x
## [1] "diabetes"
##
## $y
## [1] "weight"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=bp.1s))+geom_boxplot()
## Warning: Removed 3 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="bp.1s")
## $x
## [1] "diabetes"
##
## $y
## [1] "bp.1s"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=bp.1d))+geom_boxplot()
## Warning: Removed 3 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="bp.1d")
## $x
## [1] "diabetes"
##
## $y
## [1] "bp.1d"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=bp.2s))+geom_boxplot()
## Warning: Removed 122 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="bp.2s")
## $x
## [1] "diabetes"
##
## $y
## [1] "bp.2s"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=bp.2d))+geom_boxplot()
## Warning: Removed 122 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="bp.2d")
## $x
## [1] "diabetes"
##
## $y
## [1] "bp.2d"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=waist))+geom_boxplot()
## Warning: Removed 1 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="waist")
## $x
## [1] "diabetes"
##
## $y
## [1] "waist"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=hip))+geom_boxplot()
## Warning: Removed 1 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="hip")
## $x
## [1] "diabetes"
##
## $y
## [1] "hip"
##
## attr(,"class")
## [1] "labels"
ggplot2::ggplot(df1, aes(x=diabetes, y=time.ppn))+geom_boxplot()
## Warning: Removed 1 rows containing non-finite values (`stat_boxplot()`).
labs(x="diabetes", y="time.ppn")
## $x
## [1] "diabetes"
##
## $y
## [1] "time.ppn"
##
## attr(,"class")
## [1] "labels"
library(MASS) ##for lda function
library(klaR) ##for partimat function to produce 2-D partition plots
library(ICS) ##for multivariate normality tests
## Loading required package: mvtnorm
set.seed(12)
## Carry out LDA on training data
lda.dia <- MASS::lda(diabetes ~ ., data=train)
## Obtain ouput from LDA
lda.dia
## Call:
## lda(diabetes ~ ., data = train)
##
## Prior probabilities of groups:
## 0 1
## 0.8169014 0.1830986
##
## Group means:
## id chol stab.glu hdl ratio glyhb locationLouisa
## 0 18956.03 211.0000 94.51724 52.77586 4.360345 4.698793 0.6379310
## 1 14739.62 269.9231 194.53846 44.07692 6.669231 10.166154 0.4615385
## age gendermale height weight framelarge framemedium framesmall
## 0 47.48276 0.4827586 65.82759 182.3621 0.1896552 0.5517241 0.2068966
## 1 61.84615 0.4615385 66.69231 190.6923 0.3846154 0.5384615 0.0000000
## bp.1s bp.1d bp.2s bp.2d waist hip time.ppn
## 0 156.6034 96.18966 154.0517 93.84483 38.72414 43.62069 249.1379
## 1 163.5385 92.07692 158.3846 91.07692 41.15385 44.76923 310.3846
##
## Coefficients of linear discriminants:
## LD1
## id -0.0000131922
## chol 0.0062459949
## stab.glu -0.0102640134
## hdl -0.0311011385
## ratio -0.3959947686
## glyhb 1.0892613534
## locationLouisa 0.4820219485
## age 0.0029934818
## gendermale -0.3476056436
## height 0.0553756283
## weight -0.0098487545
## framelarge -0.7259419598
## framemedium -1.5583054564
## framesmall -1.5330626179
## bp.1s 0.0049969093
## bp.1d 0.0149080103
## bp.2s -0.0030227078
## bp.2d -0.0146735252
## waist 0.0478996942
## hip 0.0028482749
## time.ppn 0.0007788277