Add support for log-normal distribution

[mutlusun]

First of all, many thanks for all the work on this awesome package!

Please briefly describe your problem and what output you expect.

I would like to create / simulate a log-normal distribution, however the package lacks this function at the moment. There are probably some ways to work around this, but an official support would be nice!

Thanks!

[assignUser]

Thank you for the suggestion! I am currently thinking about restructuring the way we handle different distributions to make maintenance easier and better support new and custom distributions (@kgoldfeld FYI :D).

Until then the easiest way would be to use the non-random distribution:

library(simstudy)

def <- defData(
    varname = "log_normal",
    formula = "rlnorm(..n, 3, 1)",
    dist = "nonrandom"
)

n <- 10
genData(n, def)
#>     id log_normal
#>  1:  1   47.49277
#>  2:  2   27.02211
#>  3:  3   29.59762
#>  4:  4  137.59722
#>  5:  5   20.31148
#>  6:  6   31.99283
#>  7:  7   11.32037
#>  8:  8   32.01089
#>  9:  9   19.46739
#> 10: 10   12.42206

^{Created on 2021-06-19 by the reprex package (v0.3.0.9001)}

It is important to pass n as a double-dot var as the formula does not have access to it otherwise, so just define it explicitly before calling genData.

You can of course also use other variables defined in def for mean and sd, because rlnorm is not vectorized we have to use mapply. Just replace n with one in that case so the formula is run for each row separately:

library(simstudy)

def <- defData(varname = "mean", formula = 1, variance = .5, dist = "normal") 
def <- defData(def, varname = "sd", formula = 1, dist = "nonrandom") 
def <- defData(def,
        varname = "log_normal",
        formula = "mapply(..rlnorm, meanlog = mean, sdlog = sd, list(n = 1))",
        dist = "nonrandom"
    )

genData(10, def)
#>     id       mean sd log_normal
#>  1:  1 -0.2031097  1  0.1046788
#>  2:  2  0.7177785  1  3.2165236
#>  3:  3  1.4812141  1  0.8749564
#>  4:  4  1.4459571  1  1.1580688
#>  5:  5  0.8243017  1  3.4912104
#>  6:  6  0.7569261  1  0.2636364
#>  7:  7  1.1386356  1 10.2438622
#>  8:  8  2.1595046  1 46.3078970
#>  9:  9  2.2524009  1 41.5428683
#> 10: 10  2.3570354  1 55.5207466

^{Created on 2021-06-19 by the reprex package (v0.3.0.9001)}

[mutlusun]

Thanks for your fast reply and the example! :)

I have now seen that in the vignette custom functions are already described 🤷

[assignUser]

:) As this workaround seems to work for you I will close this issue. If you have any more questions don't hesitate to comment/reopen! The custom distributions support etc. will likely be tracked in #75 or mentioned there if you want to follow up.

[assignUser]

@mutlusun You could of course also make it a bit easier to read using Vectorize:

library(simstudy)

log_norm <- Vectorize(rlnorm, c("meanlog", "sdlog"))

def <- defData(varname = "mean", formula = 1, variance = .5, dist = "normal") 
def <- defData(def, varname = "sd", formula = 1, dist = "nonrandom") 
def <- defData(def,
        varname = "log_normal",
        formula = "log_norm(1, mean, sd)",
        dist = "nonrandom"
    )

genData(10, def)
#>     id       mean sd log_normal
#>  1:  1  1.0137600  1  9.8496741
#>  2:  2  1.6038807  1  6.0124375
#>  3:  3 -0.6702917  1  1.4655686
#>  4:  4  0.2745950  1  1.6558383
#>  5:  5  0.5995254  1  3.6522213
#>  6:  6  1.5485464  1  6.2918779
#>  7:  7  1.3193654  1 12.4443602
#>  8:  8  0.8407032  1  0.7066864
#>  9:  9 -0.2637986  1  0.5728164
#> 10: 10  0.6323783  1  3.6803245

^{Created on 2021-06-22 by the reprex package (v2.0.0)}

[mutlusun]

Thanks @assignUser for your detailed help! I will look into this!