Categorical variable coded as integer fails

[edwindj]

When a categorical variable is coded as an integer, the resulting mip solver, does not know that it is categorical.

library(errorlocate)

## Loading required package: validate

rules <- validator(if (sector %in% c(1,2)) turnover > 0)

# faulty record
data <- data.frame(sector = 1, turnover = 0)
weight <- c(sector = 2, turnover = 1)

# no errors found!
locate_errors(data, rules, weight=weight)$errors

##      sector turnover
## [1,]  FALSE    FALSE

This works though:

library(errorlocate)
rules <- validator(if (sector %in% c("1","2")) turnover > 0)

# faulty record
data <- data.frame(sector = "1", turnover = 0)
weight <- c(sector = 2, turnover = 1)

# turnover is faulty
locate_errors(data, rules, weight=weight)$errors

##      sector turnover
## [1,]  FALSE     TRUE

Thanks to Jeffrey Hoogland for reporting

[edwindj]

It now creates a warning:

library(errorlocate)

## Loading required package: validate

rules <- validator(if (sector %in% c(1,2)) turnover > 0)

# faulty record
data <- data.frame(sector = 1, turnover = 0)
weight <- c(sector = 2, turnover = 1)

# no errors found, but a warning is given
locate_errors(data, rules, weight=weight)$errors

## Warning: 'sector' seems a categorical variable, please recode it as a factor in the data.
## Only use character or logical values in %in% statements to prevent this warning.

##      sector turnover
## [1,]  FALSE    FALSE

Recoding sector as factor fixes the issue:

data$sector <- factor(data$sector)
locate_errors(data, rules, weight=weight)$errors

## Warning: 'sector' seems a categorical variable, please recode it as a factor in the data.
## Only use character or logical values in %in% statements to prevent this warning.

##      sector turnover
## [1,]  FALSE     TRUE