R glmnet : "(list) object cannot be coerced to type 'double' "

Question

I'm trying to use the glmnet package on a dataset. I'm using cv.glmnet() to get a lambda value for glmnet(). Here's the dataset and error message:

> head(t2)   X1 X2        X3 X4 X5         X6    X7 X8 X9 X10 X11 X12 1  1  1 0.7661266 45  2 0.80298213  9120 13  0   6   0   2 2  2  0 0.9571510 40  0 0.12187620  2600  4  0   0   0   1 3  3  0 0.6581801 38  1 0.08511338  3042  2  1   0   0   0 4  4  0 0.2338098 30  0 0.03604968  3300  5  0   0   0   0 5  5  0 0.9072394 49  1 0.02492570 63588  7  0   1   0   0 6  6  0 0.2131787 74  0 0.37560697  3500  3  0   1   0   1 > str(t2) 'data.frame':   150000 obs. of  12 variables:  $ X1 : int  1 2 3 4 5 6 7 8 9 10 ...  $ X2 : int  1 0 0 0 0 0 0 0 0 0 ...  $ X3 : num  0.766 0.957 0.658 0.234 0.907 ...  $ X4 : int  45 40 38 30 49 74 57 39 27 57 ...  $ X5 : int  2 0 1 0 1 0 0 0 0 0 ...  $ X6 : num  0.803 0.1219 0.0851 0.036 0.0249 ...  $ X7 : int  9120 2600 3042 3300 63588 3500 NA 3500 NA 23684 ...  $ X8 : int  13 4 2 5 7 3 8 8 2 9 ...  $ X9 : int  0 0 1 0 0 0 0 0 0 0 ...  $ X10: int  6 0 0 0 1 1 3 0 0 4 ...  $ X11: int  0 0 0 0 0 0 0 0 0 0 ...  $ X12: int  2 1 0 0 0 1 0 0 NA 2 ... > cv1 <- cv.glmnet(t2[,-c(1,2,7,12)], t2[,2], family="multinomial") Error in lognet(x, is.sparse, ix, jx, y, weights, offset, alpha, nobs,  :    (list) object cannot be coerced to type 'double' 

I'm excluding columns 1,2,7,12 as they are: id column, response column, contain NA's, and contain NA's. Any suggestions would be great.

Answer 1

cv.glmnet expects a matrix of predictors, not a data frame. Generally you can obtain this via

X <- model.matrix(<formula>, data=<data>) 

but in your case, you can probably get there more easily with

X <- as.matrix(t2[,-c(1,2,7,12)]) 

since you don't appear to have any factor variables or other issues that might complicate matters.

---

Since this answer is getting plenty of hits: the glmnetUtils package provides a formula-based interface to glmnet, like that used for most R modelling functions. It includes methods for glmnet and cv.glmnet, as well as a new cva.glmnet function to do crossvalidation for both alpha and lambda.

The above would become

cv.glmnet(X2 ~ ., data=t2[-1], family="multinomial") 

NA's are handled automatically, so you don't have to exclude columns with missing values.