cci.rmd

---
title: "logistic regression 실습"
author: "Hyungseok Lee"
date: "`r Sys.Date()`"
output:
  html_document: 
    toc: yes
editor_options: 
  chunk_output_type: console
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```
## introduction

>A. "cci.csv" file 
comorbidities and 1 year mortality in 100 patients  
>B. Covariables
* patientID (char3): patients ID  
* sex (char1): M0, F1  
* agegroup (char2): 0 year = 0, 0-4 year = 1, 5~9 year = 2, 10~14 year = 3, ... over 85 = 18  
* DM1 (char1): DM without complications, yes 1, no 0  
* DM2 (char1): DM with complications, yes 1, no 0  
* PAD (char1): peripheral arterial disease, yes 1, no 0  
* CHF (char1): congestive heart failure, yes 1, no 0  
* OMI (char1): old myocardial infarction, yes 1, no 0  
* COPD (char1): chronic obstructive pulmonary disease, yes 1, no 0  
* liver1 (char1): liver disease, mild, yes 1, no 0  
* liver2 (char1): liver disease, severe, yes 1, no 0  
* CKD (char1): chronic kidney disease, yes 1, no 0   
* CVD (char1) cerebrovascular disease, yes 1, no 0  
* rheuma (char1): rheumatologic disease, yes 1, no 0  
* ulcer (char1): peptic ulcer disease, yes 1, no 0  
* dementia (char1): dementia, yes 1, no 0  
* mailg (char1): mailgnancy, yes 1, no 0  
* meta (char1): metastatic cancer, yes 1, no 0  
* hemi (char1): hemiplegia, paraplegia, yes 1, no 0  
* HIV (char1): AIDS, yes 1, no 0  
* death_1yr (char1): death with in 1 year 1, alive over 1 year 0

>C. baseline characteristics table  
>D. logistic regression analysis
 
## read.csv
```{r}
cci <- read.csv("cci.csv", sep=",", header = T)
dim(cci)
head(cci)  
str(cci)
```

### Table1 
```{r}
library(moonBook)
table1 <- mytable(death_1yr~.-patientsID, show.total = T, method = 3, max.ylev = 20, data=cci)
table1
```
Table 1 - sex, agegroup, malig, meta - significant in chi square test 

```{r}
library(car)

# multicollinerity check
fit <- glm(data = cci, death_1yr~.-patientsID-DM2)
sqrt(vif(fit))>1.4

# dipersion test
fit <- glm(data = cci, death_1yr~.-patientsID-DM2, family = 'binomial')
fit.od <- glm(data = cci, death_1yr~.-patientsID-DM2, family = 'quasibinomial')
pchisq(summary(fit.od)$dispersion * fit$df.residual, fit$df.residual, lower = F)

```

### logistic regression
```{r}
result_glm <- glm(data = cci, death_1yr~.-patientsID-DM2)
summary(result_glm)
```

## variable slection
```{r}
result_glm <- step(result_glm, directon ="backward")
summary(result_glm)
```
names(cci)

## backward, forward selection
```{r}
full.model <-glm(death_1yr~.-patientsID-DM2, data = cci)
reduced.model <- step(full.model, direction = "backward")

min.model <- glm(death_1yr~1, data = cci)
fwd.model <- step(min.model, direction = "forward", 
                  scope = (death_1yr ~ sex+agegroup+DM1+PAD+CHF+OMI+COPD+
                             liver1+CKD+CVD+rheuma+ulcer+dementia+malig+
                             meta+hemi+HIV))
summary(reduced.model)
summary(fwd.model)
```


```{r}
library(leaps)

leaps <- regsubsets(death_1yr ~ sex+agegroup+DM1+PAD+CHF+OMI+COPD+
                             liver1+CKD+CVD+rheuma+ulcer+dementia+malig+
                             meta+hemi+HIV, data = cci, nbest = 10)
plot(leaps, scale='adjr2', main='Adjusted R^2')
```
## top black : agegroup, OMI, dementia, malig, meta 

## logistic regression with agegroup, OMI, dementia, malig, meta
```{r}
result_1 <- glm(data = cci, death_1yr~agegroup+OMI+dementia+malig+meta)
summary(result_1)
```

```{r}
ORtable=function(x,digits=2){
    suppressMessages(a<-confint(x))
    result=data.frame(exp(coef(x)),exp(a))
    result=round(result,digits)
    result=cbind(result,round(summary(x)$coefficient[,4],3))
    colnames(result)=c("OR","2.5%","97.5%","p")
    result
}
exp(result_1$coefficients)
ORtable(result_1)
ORplot(result_1,type=2,show.OR=T,show.CI=T)
```


## check interaction between mailg and meta
```{r}
result_2 <- glm(data = cci, death_1yr~malig*meta+agegroup+OMI+dementia+malig+meta)
summary(result_2)
```