Writing new `tidy()` method

Rationale

statim is so extendable, you can extend the existing methods or write your own and make it extensive.

Here, I plan to extend TTEST() with another bootstrapping by simply adding another variation:

add_variant(TTEST, x_by, "another_boot") %<-% variant(
    fn = function(.proc, .n = 1000L) {
        x = .proc$x_data[[1]]
        group_data = .proc$group_data
        
        grp = as.character(group_data[[1]])
        lvls = unique(grp)
        x1 = x[grp == lvls[[1]]]
        x2 = x[grp == lvls[[2]]]
        boot_fn = function(d, i) mean(d[i, 1]) - mean(d[i, 2])
        b = boot::boot(data.frame(x1, x2), boot_fn, R = .n)
        boot::boot.ci(b, type = "perc")
    }
)

another_boot_tt = 
    sleep |> 
    define_model(extra %by% group) |> 
    prepare_test(TTEST) |> 
    via("another_boot") |> 
    conclude()

another_boot_tt
#> 
#> == Model ======================================================================= 
#> 
#> Variable Mapper : x_by 
#> Args : extra | group 
#>     x_vars : 1 
#>     by_vars : 1 
#> 
#> == T-Test · another_boot ======================================================= 
#> 
#> BOOTSTRAP CONFIDENCE INTERVAL CALCULATIONS
#> Based on 1000 bootstrap replicates
#> 
#> CALL : 
#> boot::boot.ci(boot.out = b, type = "perc")
#> 
#> Intervals : 
#> Level     Percentile     
#> 95%   (-2.44, -0.95 )  
#> Calculations and Intervals on Original Scale

statim has built-in output S7 classes support, which you can use them to register your own method into that specific class in order to be automatically processed into statim internals, like extracting your output in tibble format with tidy(). Otherwise, you must register them by your own.

An example above doesn’t register with one of built-in output S7 classes, so tidy() will fail. Fortunately, you can bring it down by using making_tidy(<stat_fn>, <model_id / S7::class_formula>) %<-% method_tidy(...)type of registry:

making_tidy(TTEST, x_by) %<-% method_tidy(
    another_boot = function(.x, ...) {
        dat = .x@data
        tibble::tibble(
            R = dat$R, 
            t0 = dat$t0
        )
    }
)

another_boot_tt |> tidy()
#> # A tibble: 1 × 2
#>       R    t0
#>   <int> <dbl>
#> 1  1000 -1.58

You can clean the TTEST() in x_by model pipeline slate by simply using remove_variant().

remove_variant(TTEST, x_by, "another_boot")

The art of extensibility 2

Rationale