SAS中Proc logisitc過(guò)程提供了很完善的logistic回歸的分析功能，學(xué)習(xí)R中完成此過(guò)程只是想比較一下兩個(gè)軟件在完成此過(guò)程的差別。雖然有很多帖子介紹如何采用R完成logistic回歸過(guò)程，但是都相對(duì)過(guò)于簡(jiǎn)單，對(duì)于以下常用細(xì)節(jié)很少涉及。

1、模型篩選方法

2、如何簡(jiǎn)單設(shè)定啞變量

3、針對(duì)分類(lèi)變量，如何選取特定水平作為參考水平

4、如何簡(jiǎn)單輸出OR值及置信區(qū)間

5、如何構(gòu)建條件logistic回歸過(guò)程

6、不同模型的預(yù)測(cè)效果比較

雖然之前有很多帖子比較SAS與R的差別，但是多是基于宏觀層面的。通過(guò)比較實(shí)現(xiàn)某一具體過(guò)程的細(xì)微差別，估計(jì)更能體會(huì)兩者的功能差異。

以下是自己學(xué)習(xí)的筆記，附有一些簡(jiǎn)單說(shuō)明，供參考，希望對(duì)大家有幫助

1. library(stats)

2. help(infert) # Description of data

3. infert <- data.frame(infert)

4. str(infert) # Check type of variables

5. summary(infert) # Statistical summary

6.

7. ## Model1 Develop a simple logistic regression:

1. model1 <- glm(case ~ spontaneous+induced, data = infert, family = binomial())

2.

3. ## Model output

4. summary(model1) # Output summary information

5. confint(model1) # Output 95% CI for the coefficients

6. exp(coef(model1)) # Output OR (exponentiated coefficients)

7. exp(confint(model1)) # 95% CI for exponentiated coefficients

8. predict(model1, type="risk") # predicted values

9. residuals(model1, type="deviance") # residuals

10.

11. ## Model2 Develop a logistic regression adjusted for other potential confounders:

1. summary(model1) # Output summary information

2. confint(model1) # Output 95% CI for the coefficients

3. exp(coef(model1)) # Output OR (exponentiated coefficients)

4. exp(confint(model1)) # 95% CI for exponentiated coefficients

5. predict(model1, type="risk") # predicted values

6. residuals(model1, type="deviance") # residuals

7.

8. ## Model2 Develop a logistic regression adjusted for other potential confounders:

9. summary(model4)

10.

11. ## Model5 Conditional logistic regression

12. ## Conduct a subgroup analysis

13. ## "subset" is not aviable for "clogit"

14. ## create the subset first

15.

16. str(infert$education)

17.

18. infert1 <- subset(infert,education =="12+ yrs")

19.

20. model5 <- clogit(case ~ factor(spontaneous)+ factor(induced)+ strata(stratum),

21. data = infert1)

22. summary(model5)

23.

24. ## Model6 General logistic regression