mirror of
https://github.com/ArthurDanjou/ArtStudies.git
synced 2026-01-22 11:51:33 +01:00
75 lines
1.6 KiB
Plaintext
75 lines
1.6 KiB
Plaintext
```{r}
|
|
library(rmarkdown)
|
|
health <- read.table("health.txt", header = TRUE, sep = " ", dec = ".")
|
|
paged_table(health)
|
|
```
|
|
|
|
```{r}
|
|
Health <- health[2:5]
|
|
|
|
library(dplyr)
|
|
library(corrplot)
|
|
correlation_matrix<-cor(Health)
|
|
corrplot(correlation_matrix, order = 'hclust',addrect = 3)
|
|
```
|
|
|
|
```{r}
|
|
model <- lm(y ~ ., data = Health)
|
|
|
|
coefficients(model)
|
|
summary(model)
|
|
```
|
|
|
|
```{r}
|
|
library(ggfortify)
|
|
library(car)
|
|
autoplot(model,1:3)
|
|
```
|
|
|
|
The points are not well distributed around 0 -> [P1] is not verified
|
|
The points are not well distributed around 1 -> [P2] is not verified
|
|
The QQPlot is align with the line y = x, so it is globally gaussian -> [P4] is verified
|
|
|
|
```{r}
|
|
set.seed(0)
|
|
durbinWatsonTest(model)
|
|
```
|
|
|
|
The p-value is 0.58 > 0.05 -> We do not reject H0 so the residuals are not autocorrelated -> [P3] is verified
|
|
|
|
```{r}
|
|
library(GGally)
|
|
ggpairs(Health, progress = F)
|
|
```
|
|
|
|
We observe that the variable age is correlated with the variable y. There is a quadratic relation between both variables.
|
|
|
|
```{r}
|
|
Health2 <- Health
|
|
Health2$age_sq <- Health2$age^2
|
|
Health2 <- Health2[1:24,]
|
|
|
|
model2 <- lm(y ~ ., data = Health2)
|
|
|
|
summary(model2)
|
|
coefficients(model2)
|
|
```
|
|
|
|
```{r}
|
|
library(ggfortify)
|
|
library(car)
|
|
autoplot(model2,1:4)
|
|
```
|
|
|
|
The points are well distributed around 0 -> [P1] is verified
|
|
The points are not well distributed around 1 -> [P2] is verified
|
|
The QQPlot is align with the line y = x, so it is gaussian -> [P4] is verified
|
|
|
|
```{r}
|
|
set.seed(0)
|
|
durbinWatsonTest(model2)
|
|
```
|
|
|
|
The p-value is 0.294 > 0.05 -> We do not reject H0 so the residuals are not autocorrelated -> [P3] is verified
|
|
|