For loop to call a different function variable

Question

I want call the predict function using different models to extract its predicted value. I tried using paste0 to call the right model but it doesn't work for example:

model0 = lm(mpg ~ cyl + disp, data = mtcars)
model1 = lm(mpg ~ hp + drat, data = mtcars)
model2 = lm(mpg ~ wt + qsec, data = mtcars)

testdat0 = data.frame(cyl = 6, disp = 200)
testdat1 = data.frame(hp = 100, drat = 4)
testdat2 = data.frame(wt = 4, qsec = 20)

res = NULL
for (i in 1:3) {
  res = rbind(res, c(i-1, predict(paste0('model',i-1), newdata = paste0('testdat0',i-1))))
}

to do it manually

rbind(c(0, predict(model0, newdata = testdat0)), 
      c(1, predict(model1, newdata = testdat1)), 
      c(2, predict(model2, newdata = testdat2)))

              1
[1,] 0 21.02061
[2,] 1 24.40383
[3,] 2 18.13825

Another way I thought of doing this was to put the models and testdata in 2 separate list() and use a for loop to call them but that also didn't work. Is there another way of doing this or am I doing something wrong.. TIA

the predict you do only extracts the value for the first category (in your example the Mazda RX4) since you c() it to an index number. Is this what you want to do? — Julian_Hn
– Julian_Hn, Commented Apr 29, 2019 at 6:31
Maybe: lm<-list(model0,model1,model2);lt<-list(testdat0,testdat1,testdat2);purrr::map2(lm,lt,predict) or mapply(predict, lm,lt) — A. Suliman
– A. Suliman, Commented Apr 29, 2019 at 6:42
Yes it was intentional. This is just a silly example showing model comparison to predict mpg — Aystealthy
– Aystealthy, Commented Apr 29, 2019 at 6:58

Julian_Hn · Accepted Answer · 2019-04-29 07:05:05Z

2

My solution to your problem using a list and sapply so we don't need to define an external variable and rbind() to it over and over.

model0 = lm(mpg ~ cyl + disp, data = mtcars)
model1 = lm(mpg ~ hp + drat, data = mtcars)
model2 = lm(mpg ~ wt + qsec, data = mtcars)

testdat0 = data.frame(cyl = 6, disp = 200)
testdat1 = data.frame(hp = 100, drat = 4)
testdat2 = data.frame(wt = 4, qsec = 20)

#make list from sample data
data <- list(dat0=list(model=model0,test=testdat0),
             dat1=list(model=model1,test=testdat1),
             dat2=list(model=model2,test=testdat2))

#sapply over list, automatically converts to matrix
res <- sapply(data,function(dat) predict(dat$model,newdata=dat$test) )

> res
  dat0   dat1   dat2 
21.02061 24.40383 18.13825

edited Apr 29, 2019 at 7:05

answered Apr 29, 2019 at 6:37

Julian_Hn

2,1411 gold badge10 silver badges19 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

AkselA Over a year ago

This is certainly a much better method than using rbind() over and over again.

Julian_Hn Over a year ago

@AkselA Thanks for the heads-up fixed the comment.

Aystealthy Over a year ago

sorry if this is a stupid question but I'm confused about the output of your 'res', what is it doing/showing? the expect mpg for each car based on the 3 models & data?

Julian_Hn Over a year ago

I'm sorry. I made a code error (data instead of dat) and did not notice that the result ist nonsensical (it's very early here). I fixed it and it should now show the exprected mpg for the values in testdat. // PS: To clarify what happened here: If you do not provide a newdata argument, predict will predict the original data with the model. Since data$test does not exist, R will handle the function as if newdata was omitted instead of throwing an error (I really don't like this behaviour). So it just predicted the data in mtcars with the model.

Sonny · Accepted Answer · 2019-04-29 07:29:25Z

1

To make your for loop work, you can make the below changes:

res = NULL
for (i in 1:3) {
  res = rbind(res, c(i-1, predict(eval(as.name(paste0('model',i-1)))), newdata = eval(as.name(paste0('testdat',i-1)))))
}

edited Apr 29, 2019 at 7:29

answered Apr 29, 2019 at 6:23

Sonny

3,1651 gold badge13 silver badges19 bronze badges

Collectives™ on Stack Overflow

For loop to call a different function variable

2 Answers 2

4 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

4 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related