add Missing values to specific columns

[Karim-Mane]

Is your feature request related to a problem? Please describe.
This is a follow up on the introduction of 10% of missing data.

Thanks @joshwlambert for the clarification about the usage of NA in some columns. As you mentionned, in the discussion of PR #187, boils down to represent unavailable data for cases where this is not collected (Ct values for non confirmed cases).

I have run the function several times to see columns where NA is introduced from the sim_linelist() function. It came out to be columns - see the outcome below.

final_linelist <- NULL
# simulate linelist 100 times
for (i in seq_len(100)) {
    linelist <- suppressWarnings(simulist::sim_linelist())
    x <- matrix(colSums(is.na(linelist)), ncol=ncol(linelist))
    final_linelist <- rbind(final_linelist, x)
}
colnames(final_linelist) <- names(linelist)

# detect columns with NAs
sum_na <- colSums(final_linelist)
names(sum_na) <- colnames(final_linelist)
sum_na[sum_na > 0]
#>     date_admission       date_outcome date_first_contact  date_last_contact 
#>              62589              67113                100                100 
#>           ct_value 
#>              38826

^{Created on 2025-02-19 with reprex v2.1.0}

Describe the solution you'd like
I suggest the 10% NA to be introduced in the remaining columns i.e. in columns other than these five columns.

[joshwlambert]

Thanks for the informative description. I have addressed this request in PR #199. I've added a new internal .add_missing() function that adds what you've requested.

If the missing_value is NA then the newly inserted missing values do not sample from the existing NA elements in the <data.frame>. Avoiding overwriting missing values. If the missing_value is changed by the user, for example to "N/A", then the .add_missing() function samples from all <data.frame> elements.

.add_missing() also performs type coercion to avoid unwanted type coercions when the user specifies a custom missing_value.

I suggest the 10% NA to be introduced in the remaining columns i.e. in columns other than these five columns.

The approach taken in .add_missing() still allows introducing missing values into the <data.frame> columns that already contain NAs when the missing_value = NA (default), this nicely retains the feature of random missingness without overwriting NA values.