0

I'm trying to take Lasso regression but I can't define well my X and Y in R.

#load data

>test.data<-read.spss("C:\\Users\\Inhib\\OneDrive\\documents\\dummy.sav",use.value.labels=TRUE, to.data.frame=TRUE)
>test.data #testing my data, it's all there so I won't add it here
#take columns 2 to 6 for X
>X<-as.matrix(test.data[,2:6])


# Column 1 is the predicted variable Y


>Y<-as.matrix(test.data[,1])


 #Ok, let's fit it 
fit<- glmnet(x, y, family="gaussian", alpha=0, lambda=0.001)

then I get this error message: 

"Error in glmnet(x, y, family = "gaussian", alpha = 0, lambda = 0.001) : 
  number of observations in y (100) not equal to the number of rows of x   (222)"

Now all columns are of the same length (222) but the error says that there are only 100 in Y and 222 in X.

#So I checked for Y here

>length(Y)
[1] 222
#Then checked for X
> length(X)
[1] 1110

Now, it certainly points to that I miss something. Obviously the matrix size is different and X is 222*5 columns, but how can I make it work? Meaning that this error will be gone...I tried many ways. I'm working on that for hours, that really stop me from progressing and there is not much help from google. Would be grateful for solution.

Gilad Sabo
  • 41
  • 1
  • 5
  • What are `x` and `y` (lower case) you're using in `glmnet` ? try running `rm(list=ls())` and running your whole code again. – Tensibai Jul 13 '16 at 13:40

1 Answers1

0

I would recommend
1) Check the dimension of matrix using dim(x) and dim(y) rather than length
2) Check if there is any NA value in y (highly possible)

DevMui
  • 101
  • 4
  • dim(x) gave me [1]222 4, dim(y) gave me NULL. Also, I checked and there are not NA values in the data. How can it be solved? – Gilad Sabo Jul 13 '16 at 13:46
  • It does not feel right as dim(x) should give 222 5 and dim(y) should give 222 1. I guess you did not capitalize X and Y in glmnet(x, y, family="gaussian", alpha=0, lambda=0.001) which might be problem? Shouldn't it be glmnet(X, Y, family="gaussian", alpha=0, lambda=0.001)? – DevMui Jul 13 '16 at 13:54