Linear regression loop with data.table; "Error in data.table column or argument (nr) is NULL"

Asked Aug 23 '18 at 15:44

Active Sep 01 '18 at 14:19

Viewed 254 times

As my dataset is cumbersomely large, I would like to automate some procedures. I found this link, which proposes a linear regression loop, which for the dataset mtcars is as follows:

data.table(mtcars)[, .(MyFits = lapply(.SD, function(x) if(is.numeric(x)) summary(lm(mpg ~ x)))), .SDcols = -1]

I have tried to apply this onto my own dataset with limited succes. I do get the output but there is a problem. The result for some of the Fits is NULL, so when I try to do the suggested operation Fits[, lapply(MyFits, coef)] I get:

Error in data.table column or argument 3 is NULL

Can I somehow make Fits[, lapply(MyFits, coef)]skip the MyFits which are NULL?

EDIT: Question to which the comments refers has been removed for clarity.

edited Sep 01 '18 at 14:19

asked Aug 23 '18 at 15:44

Tom

2,173
1
17
44

2

`Fits = data.table(mtcars)[, .(nm = names(.SD), MyFits = lapply(.SD, function(x) if(is.numeric(x)) summary(lm(mpg ~ x)))), .SDcols = -1]`? – Frank Aug 23 '18 at 15:53
Always very nice when people down vote a question without saying why.. – Tom Aug 30 '18 at 05:46
1

@Frank, Your answer worked by the way, thank you very much! – Tom Aug 30 '18 at 05:46
Dear Tom, Please fill free to post an answer and accept it if it works. – Artem Sep 16 '18 at 20:47
@Frank post as answer, please? – zx8754 Feb 28 '19 at 13:39

Linear regression loop with data.table; "Error in data.table column or argument (nr) is NULL"

0 Answers0

Linked