lhz1029 / faker Goto Github PK

My original dataset has columns which are not normally distributed but, when I simulate them, the fake data is normally distributed. Is there a way to simulate the data but keep the original distribution? Cheers!

The following command returns 150 rows (like the original iris dataset) and not 10 as indicated.

nrow(fakeR::simulate_dataset(iris, n = 10))
However, when the Species, the categorical variable, is removed, it works fine.

nrow(fakeR::simulate_dataset(iris[,-5], n = 10))
I believe, the problem occurs in this bit of the simulate_dataset function:

 if (unorder > 0) {
      print("Some unordered factors...")
      fake_categorical <- .simulate_categorical(dataset, location, row)
    }

When you try to simulate a dataset with an ordered factor and stealth.level=1 it throws an error that there is a non-numeric argument to binary operator

> df<-mtcars
> df$carb<-as.factor(df$carb)
> df$gear<-factor(df$gear,ordered=TRUE)
> sim_df<-simulate_dataset(dataset = df,stealth.level = 1)
[1] "Some unordered factors..."
[1] "Some numeric variables and ordered factors..."
Error in n * ncol(sigma) : non-numeric argument to binary operator

Instead of simulating the dataset and "taking into account the covariances between the numeric and ordered factors"

I saw that fakeR isn't on CRAN any longer, but I feel like it deserves to be.

Would you be interested in getting it past the CRAN checks? If not, I'd be happy to help.

Thanks

Steve