I tried to split the data(bank) into training data and test data. But I somehow got an error below.How can I solve this problem?
train = bank[1:100, ] test = bank[!train,] Status.test =Status[!train] glm.fit=glm(Status~Length+Right+Bottom+Top+Diagonal,data=bank,family=binomial,subset=train) #Error in xj[i] : invalid subscript type 'list' glm.probs=predict(glm.fit,test,type="response") glm.pred=rep("genuine",100) glm.pred[glm.probs>.5]="counterfeit" table(glm.pred,test)##classification on training data #Error in table(glm.pred, test) : all arguments must have the same length 3 Answers
The issue is in subset=train. According to the ?glm. the subset should be a vector as oppose to a subset of original dataset:
subset an optional vector specifying a subset of observations to be used in the fitting process.
Hence, you may need to change the code to: glm.fit=glm(Status~Length+Right+Bottom+Top+Diagonal,data=train,family=binomial)
or
glm.fit=glm(Status~Length+Right+Bottom+Top+Diagonal,data=bank,family=binomial,subset=1:100)
Generally, you could achieve what you asked by doing something like this: Assume column 'response' is observed column:
samples=1:100 train = bank[samples, ] test = bank[-samples,] Status.test =bank[samples,'response'] BTW, I would suggest using sample() function in order to take samples randomly for train and test. like this:
samples=sample(nrow(bank), 0.8*nrow(bank)) train = bank[samples, ] test = bank[-samples,] Status.test =bank[samples,'response'] If you set the training data like:
data[1: 100,] Then in lm() function you use the argument:
data = bank[train,] Alternatively you can set train like:
seq(1: 100) as a sequence of indices, you need to use in the
lm(): data = bank, subset = train 
