Implement method to return a pretty table about rsimsum HOT 6 OPEN

Hello, thanks for creating such an amazing package for summarizing simulation results. I saw this issue so I want to add a small suggestion here. It seems the print.summary.simsum() function will only print those nice tables but the summary tables are not returned. This is somewhat inconvenient for channeling the R output for other uses. I know I can extract the "tidy" long table from the summary object directly or use the tidy() function but they are not good for presentation purposes. What are your thoughts on this? Thank you very much!

from rsimsum.

Hi, thanks for your feedback!
Just to understand better your comment, would you like print.summary.simsum() to return the formatted tables that are printed to the R console?

from rsimsum.

Yes. I'll be more specific. Allow me to explain.

Currently the print.summary.simsum() doesn't return anything other than printing the required summary tables. Personally, I think a better design is that it returns a (list) of summary data frames, maybe the output object in your code. I want this feature because such tables are better suited for presenting the results than the tidy format summary tables if I have a methodvar. And I cannot really use R's printed data frame conveniently for other purposes, e.g., making it into kables.

  ### If methodvar, put them side by side
  if (!is.null(x$methodvar)) {
    x$summ <- .bind_methods(data = x$summ, by = x$by, methodvar = x$methodvar)
  }

  ### Split by summary statistics for printing
  x$summ[["Performance Measure"]] <- droplevels(x$summ[["Performance Measure"]])
  output <- .split_by(data = x$summ, by = "Performance Measure")

  ### Print
  for (i in seq_along(output)) {
    cat(paste0("\n", names(output)[i], ":\n"))
    output[[i]][["Performance Measure"]] <- NULL
    print(output[[i]], row.names = FALSE)
  }

from rsimsum.

Aha, now I see – something like the example below?

library(rsimsum)
s2 <- simsum(data = relhaz, estvarname = "theta", true = -0.50, se = "se", methodvar = "model", by = c("baseline", "n"))
#> 'ref' method was not specified, Cox set as the reference
a <- print(summary(s2, stats = "bias"))
#> Values are:
#>  Point Estimate (Monte Carlo Standard Error)
#> 
#> Bias in point estimate:
#>     baseline   n              Cox              Exp            RP(2)
#>  Exponential  50  0.0215 (0.0328)  0.0239 (0.0326)  0.0183 (0.0331)
#>  Exponential 250 -0.0215 (0.0149) -0.0214 (0.0151) -0.0227 (0.0149)
#>      Weibull  50 -0.0282 (0.0311)  0.1509 (0.0204) -0.0348 (0.0311)
#>      Weibull 250 -0.0120 (0.0133)  0.1482 (0.0093) -0.0139 (0.0137)
library(knitr)
#> 
#> Attaching package: 'knitr'
#> The following object is masked from 'package:rsimsum':
#> 
#>     kable
kable(a[[1]], caption = names(a)[1], align = "r")

Bias in point estimate

^{Created on 2021-09-16 by the reprex package (v2.0.1)}

from rsimsum.

Yes, that's fantastic!

from rsimsum.

Awesome! I just pushed this updated version to GitHub, you can try it out with:

library(remotes)
remotes::install_github("ellessenne/github")

Let me know your thoughts it you get the chance to play around with it! 😃

from rsimsum.