I’m analyzing count data (insect density in maize crop) and performing negative binomial regression analysis using 30 predictors (remotely-sensed variables) to predict number of insects. I was trying to execute a leave one-out cross-validation method to test the performance of the models using R software (the cv.glm function in boot package). Seemingly, the cross-validation method does not work for my data and I got an error message says “Error in model.frame.default(formula = data$Density ~ ., data = list(Density = c(6L, :
variable lengths differ (found for 'B1')”
· Density is the insect numbers
· B1 is predictor number 1.
In fact, the data for the predictor (B1) and Density are of the same size.
I appreciate any suggestions.
Hello Elfatih
The problem could be an error in the same size, even though you check that are the same size, R software doesn't understand this, you need to check the data origin format and the way to import to R.
Best regards.
Elfaith, cross validation is a futile exercise, which provides no additional useful information than what the full data already provides. "Independent validation", not cross validation, is the only best way to validate any model.
Hi Elfatih,
this is not a cross validation problem. Check for NAs in your data. Two vectors (variables) can have the same length but during the modelling process NAs may be removed resulting in one vector being shorted than the other one.
Also try to provide your data set in the form of a data frame not a list, in a data frame all variables must have the same length, however, this is not required in a list.
Martin
Dear all,
Thank you so much for your very useful comments and suggestions
Regards
Elfatih
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Agronomy
- Crop Science