sum() on Boolean vector not working as expected in Summarise function about simdesign

[philchalmers]

Whenever you use the c() operator R will coerce the provided elements to the most general type. So, c(1.2, 1, TRUE), will change TRUE to the value 1.0. Consider what happens when you pass this c(1,2,'a').

If you want to keep the types constant return a data.frame() in your Analyse() step. That will allow the column elements to be of different type. Just change

ret <- c(coefficientInt=coef(summary(fit))[4,1], singular=singular)

to

ret <- data.frame(coefficientInt=coef(summary(fit))[4,1], singular=singular)

[kneijenhuijs]

Thanks, I feel silly for not catching that. However, I still run into issues when including a Boolean vector in the returned data frame from the Analyse function. When I implement your suggestion, I run into the following error code and traceback:

 Error in FUN(X[[i]], ...) : 
  only defined on a data frame with all numeric variables 
7.
stop("only defined on a data frame with all numeric variables") 
6.
FUN(X[[i]], ...) 
5.
lapply(args, function(x) {
    x <- as.matrix(x)
    if (!is.numeric(x) && !is.complex(x)) 
        stop("only defined on a data frame with all numeric variables") ... 
4.
Summary.data.frame(structure(list(singular = c(FALSE, FALSE, 
TRUE, TRUE, FALSE)), row.names = c(NA, -5L), class = "data.frame"), 
    na.rm = TRUE) 
3.
Functions$summarise(results = results, condition = condition, 
    fixed_objects = fixed_objects) 
2.
Analysis(Functions = Functions, condition = design[i, ], replications = replications, 
    fixed_objects = fixed_objects, cl = cl, MPI = MPI, seed = seed, 
    bootSE = bootSE, boot_draws = boot_draws, save = save, save_results = save_results, 
    save_results_out_rootdir = out_rootdir, save_results_dirname = save_results_dirname,  ... 
1.
runSimulation(Design, replications = 5, verbose = FALSE, parallel = FALSE, 
    generate = Generate, analyse = Analyse, summarise = Summarise, 
    progress = TRUE) 

The issue seems to occur with step 5, which I assume is part of the Summary.data.frame type. I have fixed my issue for now by transforming my boolean into a vector of zeroes and ones (so that I can still easily sum up the counts of TRUE), but it seems that there are issues with returning values that are not numeric.

[philchalmers]

In your runSimulation() call, try using the edit option for debugging. This will step into the respective function so you can play around with the state of the objects directly so that you can see what's going on. When I do this:

runSimulation(..., edit = 'summarise')

I see that

Browse[2] > str(results)
'data.frame':	5 obs. of  2 variables:
 $ coefficientInt: num  1.04 -0.585 0.803 -0.383 -0.327
 $ singular      : logi  FALSE FALSE FALSE FALSE FALSE

So, results is a data.frame object with 2 columns. Running the next line bias(results[1]) doesn't work because data.frames should be indexed either by column (results[,1]), the $ sign (results$coefficients), or lists (results[[1]]). Therefore, all you need to change is the [#] components to [[#]] to fix your code. HTH.