trtAssign converts all columns to integer

[hebaalhosainyTU]

[hebaalhosainyTU]

After applying the trtAssign function, the values in randomEffect column changed and not anymore normally distributed.

[assignUser]

Hello @hebaalhosainyTU thanks for the report. Next time please consider using reprex (see below) as it makes it easier to reproduce issues.

@kgoldfeld Reprex below is rendered with cran version of simstudy and indeed looks problematic. I haven't had time to look deeoer into it though.

library(simstudy)
ydef <- defData(varname = "randomEffect", dist = "normal", formula = 0, variance = sqrt(3))

indData <- genData(24, ydef)
head(indData)
#>    id randomEffect
#> 1:  1    2.9851502
#> 2:  2    0.9825252
#> 3:  3   -0.3632496
#> 4:  4   -0.3291816
#> 5:  5    0.4210097
#> 6:  6    1.5862488
indData <- trtAssign(indData, nTrt = 2, balanced = TRUE, grpName = "T")
head(indData)
#>    id randomEffect T
#> 1:  1            2 1
#> 2:  2            0 1
#> 3:  3            0 1
#> 4:  4            0 1
#> 5:  5            0 1
#> 6:  6            1 0

^{Created on 2022-09-28 with reprex v2.0.2}

[assignUser]

Running the code a few times shows that the randomEffect col is rounded to 0 decimal places...

[assignUser]

https://github.com/kgoldfeld/simstudy/blob/main/R/group_data.R#L401-L403
Converts ALL columns to integer ⚠️

[kgoldfeld]

Yes - that is indeed a bug. If you want a quick workaround while that is fixed, you can do this:

ydef <- defData(varname = "randomEffect", 
                dist = "normal", formula = 0, variance = sqrt(10))
ydef <- defData(ydef, varname = "T", 
                dist = "trtAssign", formula = "1;1")

genData(6, ydef)
#>    id randomEffect T
#> 1:  1    2.6744327 0
#> 2:  2    0.3138995 1
#> 3:  3   -0.9968553 0
#> 4:  4    0.2780552 1
#> 5:  5   -1.3009989 1
#> 6:  6    1.5663468 0

But great catch, I will certainly fix that bug.

[kgoldfeld]

@assignUser It is really strange how those lines of code got in there - must have been added to facilitate testing. And yet, testing did not pick that up. It is particularly odd since the output from trtAssign (really trtObserve) is already integer. Not sure why it would have seemed like a good idea to convert even the treament variable to integer. I will remove the offending code.

[kgoldfeld]

@hebaalhosainyTU This has been fixed in the current development version. Thanks again.