CODEWITHSUNDEEP
rlogistic-regressionglmnet
AME
AME

Reputation: 339

glmnet error - non-conformable arguments

I keep running into a stubborn error trying to use the glmnet stats package as indicated below.

I've tried the limited suggestions listed here (including setting the data as a data.matrix). I've also attempted to use the "penalty.box" setting described in ?glmnet with no positive results. 

df = structure(list(term = c(0, 1, 0, 0, 1, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0), int_rate = c(10.65, 15.27, 15.96, 13.49, 12.69, 7.9, 15.96, 18.64, 21.28, 12.69, 14.65, 12.69, 13.49, 9.91, 10.65, 16.29, 15.27, 6.03, 11.71, 6.03, 15.27, 12.42, 11.71, 11.71, 11.71, 9.91, 16.77, 11.71, 11.71, 7.51, 7.9, 15.96, 8.9, 15.96, 10.65, 9.91, 7.9, 12.42, 12.69, 7.51, 7.9, 18.25, 16.77, 6.03, 9.91, 8.9, 10.65, 6.03, 6.62, 9.91), emp_length = c(NA, 1, NA, NA, 1, 3, 8, 9, 4, 1, 5, NA, 1, 3, 3, 1, 4, NA, 1, 6, 3, NA, NA, 5, 1, 2, 2, NA, 1, 7, 5, 2, 2, 7, NA, 2, 1, 1, 1, 4, NA, 9, NA, NA, 6, NA, 6, NA, 5, 8), annual_inc = c(24000, 30000, 12252, 49200, 80000, 36000, 47004, 48000, 40000, 15000, 72000, 75000, 30000, 15000, 1e+05, 28000, 42000, 110000, 84000, 77385.19, 43370, 105000, 50000, 50000, 76000, 92000, 50004, 106000, 25000, 17108, 75000, 29120, 24044, 34000, 41000, 55596, 45000, 36852, 27000, 68004, 62300, 65000, 55000, 45600, 0000, 1e+05, 27000, 60000, 70000, 80000), delinq_2yrs = c(0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 3L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), inq_last_6mths = c(1L, 5L, 2L, 1L, 0L, 3L, 1L, 2L, 2L, 0L, 2L, 0L, 1L, 2L, 2L, 1L, 2L, 0L, 0L, 0L, 3L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 1L, 2L, 0L, 0L, 1L, 3L, 0L, 0L, 0L, 0L, 1L, 1L, 0L, 0L, 1L, 0L, 2L), outcome = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)), .Names = c("term", "int_rate", "emp_length", "annual_inc", "delinq_2yrs", "inq_last_6mths", "outcome"), row.names = c(NA, 50L), class = "data.frame")
X = select(df, -outcome)
Y = df$outcome
X_train = as.matrix(X[1:50,])
Y_train = as.matrix(Y[1:50])

library(glmnet)

model = glmnet(X_train, Y_train, family = "binomial")

summary(model)

Here's the error:

Error in drop(y %*% rep(1, nc)) : 
  error in evaluating the argument 'x' in selecting a method for     
function 'drop': Error in y %*% rep(1, nc) : non-conformable arguments

The actual dataset is 110 variables and ~1mm observations , but the partial dataset above is producing the same issue.

Any suggestions on debugging approaches for this?

Upvotes: 1

Views: 10068

Answers (1)

Berk U.
Berk U.

Reputation: 7188

The issue is that you have NA entries is the emp_length column, which you cannot include in the dataset that you pass to glmnet. You either need to drop rows that contain NA values from your dataset, or impute them.

Here is some code that you can use to drop the values and get glmnet to work properly.

X = select(df, -outcome)
Y = df$outcome
X_train = as.matrix(X[1:50,])
Y_train = as.matrix(sample(0:1,50,replace=TRUE))

has_NA = apply(is.na(X_train), 1, any) #= 1 if any column in that row is NA
X_train = X_train[!has_NA,]
Y_train = Y_train[!has_NA,]

library(glmnet)    
model = glmnet(X_train, Y_train, family = "binomial")

Upvotes: 1

Related Questions