CODEWITHSUNDEEP
routput
pyll
pyll

Reputation: 1764

Exporting Output from Model Summary in R

Is there a way to export the output from a summary procedure in R (similar to ods output in SAS)?

I have this time series:

summary(ets(ts(c(100, 110, 120, 200, 300, 100, 200, 100, 120, 300), start = c(2009, 1), frequency = 12)))

And when you run this you get all kinds of great information like MAPE and RMSE. I want to selectively assign some of these metrics to variables for use in other processes further down the line.

MAPE.ets <- [some bit of code that takes the MAPE from the output and pops it right here]

Is there a way in R to do this?

EDIT: I realized after some answers that stlf ets behaves differently than ets, and I need a solution to work for stlf ets as well.

summary(stlf(ts(c(100, 110, 120, 200, 300, 100, 200, 100, 120, 300), start = c(2009, 1), frequency = 4)))

Upvotes: 0

Views: 347

Answers (2)

Pierre Lapointe
Pierre Lapointe

Reputation: 16277

The summary gives you a vector. You can get MAPE like this if you want colnames and rownames.

out <-summary(ets(ts(c(100, 110, 120, 200, 300, 100, 200, 100, 120, 300), 
              start = c(2009, 1), frequency = 12)))
out[,5,drop=FALSE]

             MAPE
Training set 44.51514

I you only want the number, remove drop=FALSE

out[5]
[1] 44.51514

EDIT Using stfl, it is better to use the accuracy function instead of summary. To get MAPE, you can do:

res <- stlf(ts(c(100, 110, 120, 200, 300, 100, 200, 100, 120, 300), 
            start = c(2009, 1), frequency = 4))
accuracy(res)[5]
[1] 44.5994

Upvotes: 1

David Heckmann
David Heckmann

Reputation: 2939

simply assign the summary to a new object. This creates a matrix in which the metrics you mentioned can be accessed:

summ <- summary(ets(ts(c(100, 110, 120, 200, 300, 100, 200, 100, 120, 300), start = c(2009, 1), frequency = 12)))

summ[,"MAPE"]
[1] 44.51514

EDIT:

The intended way in the forecast package appears to be the use of the accuracy method (the summary method for the ets-class actually calls accuracy); and this will work for both ets and stlf :

res2 <- stlf(ts(c(100, 110, 120, 200, 300, 100, 200, 100, 120, 300), start = c(2009, 1), frequency = 4))
acc <- accuracy(res2)
acc[,"MAPE"]

Upvotes: 1

Related Questions