Logistic
Regression
KEY:
ds = dataset you are currently using.
​
DV = predicted variable
​
IV = predictor variable
​
XYXY = dummy name for a variable, matrix, or data frame into which you are moving information.
Logistic regression with two predictors.
(1) run regression (2) generate probabilities (3) generate predictions (4) create table and calculate accuracy
_____________________________
(1)
XYXY <- glm(DV ~ IV1 + IV2, data=ds, family = binomial)
summary(XYXY)
​
(2)
XYXYprobabilities <- predict(XYXY, ds, type = "response")
​
(3)
XYXYprediction <- ifelse(XYXYprobabilities > .5, 1, 0)
​
(4)
table(XYXYpredition, ds$DV)
mean(XYXYprediction == ds$DV)
_____________________________
​
​
Alternate to step (4), you can use the caret package to generate a more complete set of descriptives.
NOTE: the command "confusionMatrix" requires both XYXYpredition and ds$DV to be factored (i.e., as.factor())
_____________________________
library(caret)
confusionMatrix(XYXYprediction, ds$DV)
_____________________________
​
​