Plot the boostrapped mixing proportions

Source: R/plot_boostrap_mixing_proportions.R

From the results of a call to mobster_boostrap, and the results of a call to bootstrapped_statistics, a boxplot with the mixing proportions is produced.

plot_bootstrap_mixing_proportions(
  x,
  bootstrap_results,
  bootstrap_statistics,
  colors = c(Tail = "gainsboro")
)

Arguments

x

A MOBSTER fit.
bootstrap_results

Results of a call to mobster_boostrap.
colors

If provided, these colours will be used for each cluster. If a subset of colours is provided, palette Set1 from RColorBrewer is used. By default the tail colour is provided as 'gainsboro'.
bootstrap_statisticsResults

of a call to bootstrapped_statistics.

Value

A barplot of the bootstrapped mixing proportions.

Examples

# Random small dataset
dataset = random_dataset(N = 200, seed = 123, Beta_variance_scaling = 100)
x = mobster_fit(dataset$data, auto_setup = 'FAST')

#>  [ MOBSTER fit ] 
#> 
#> ✔ Loaded input data, n = 200.
#> ❯ n = 200. Mixture with k = 1,2 Beta(s). Pareto tail: TRUE and FALSE. Output
#> clusters with π > 0.02 and n > 10.
#> ! mobster automatic setup FAST for the analysis.
#> ❯ Scoring (without parallel) 2 x 2 x 2 = 8 models by reICL.
#> 
#> [easypar] 2020-10-29 11:33:49 - Overriding parallel execution setup [FALSE] with global option : FALSE
#> 
#> 
#> ℹ MOBSTER fits completed in 8.9s.
#> 
#> ── [ MOBSTER ] My MOBSTER model n = 200 with k = 2 Beta(s) without tail ────────
#> ● Clusters: π = 52% [C1] and 48% [C2], with π > 0.
#> ✖ No tail fit.
#> 
#> ● Beta C1 [n = 102, 52%] with mean = 0.58.
#> ● Beta C2 [n = 98, 48%] with mean = 0.11.
#> ℹ Score(s): NLL = -108.12; ICL = -182.58 (-182.58), H = 1.88 (1.88). Fit
#> converged by MM in 66 steps.

# Just 5 resamples of a nonparametric bootstrap run, disabling the parallel engine
options(easypar.parallel = FALSE)
boot_results = mobster_bootstrap(x$best, n.resamples = 5, auto_setup = 'FAST')

#>  [ MOBSTER bootstrap ~ 5 resamples from nonparametric bootstrap ] 
#> 
#> ── [ MOBSTER ] My MOBSTER model n = 200 with k = 2 Beta(s) without tail ────────
#> ● Clusters: π = 52% [C1] and 48% [C2], with π > 0.
#> ✖ No tail fit.
#> 
#> ● Beta C1 [n = 102, 52%] with mean = 0.58.
#> ● Beta C2 [n = 98, 48%] with mean = 0.11.
#> ℹ Score(s): NLL = -108.12; ICL = -182.58 (-182.58), H = 1.88 (1.88). Fit
#> converged by MM in 66 steps.
#> 
#> 
ℹ Creating nonparametric bootstrap resamples
#> 
✔ Creating nonparametric bootstrap resamples ... done
#> 
#> 
#> ── Running fits ─────────────────────────────────── Might take some time ...  ──
#> [easypar] 2020-10-29 11:33:58 - Overriding parallel execution setup [TRUE] with global option : FALSE
#>  [ MOBSTER fit ] 
#> 
#> ✔ Loaded input data, n = 200.
#> ❯ n = 200. Mixture with k = 1,2 Beta(s). Pareto tail: TRUE and FALSE. Output
#> clusters with π > 0.02 and n > 10.
#> ! mobster automatic setup FAST for the analysis.
#> ❯ Scoring (without parallel) 2 x 2 x 2 = 8 models by reICL.
#> 
#> [easypar] 2020-10-29 11:33:58 - Overriding parallel execution setup [FALSE] with global option : FALSE
#> 
#> 
#> ℹ MOBSTER fits completed in 8.6s.
#> 
#> ── [ MOBSTER ] My MOBSTER model n = 200 with k = 2 Beta(s) without tail ────────
#> ● Clusters: π = 51% [C1] and 49% [C2], with π > 0.
#> ✖ No tail fit.
#> 
#> ● Beta C1 [n = 101, 51%] with mean = 0.58.
#> ● Beta C2 [n = 99, 49%] with mean = 0.11.
#> ℹ Score(s): NLL = -118.81; ICL = -204.17 (-204.17), H = 1.66 (1.66). Fit
#> converged by MM in 61 steps.
#>  [ MOBSTER fit ] 
#> 
#> ✔ Loaded input data, n = 200.
#> ❯ n = 200. Mixture with k = 1,2 Beta(s). Pareto tail: TRUE and FALSE. Output
#> clusters with π > 0.02 and n > 10.
#> ! mobster automatic setup FAST for the analysis.
#> ❯ Scoring (without parallel) 2 x 2 x 2 = 8 models by reICL.
#> 
#> [easypar] 2020-10-29 11:34:07 - Overriding parallel execution setup [FALSE] with global option : FALSE
#> 
#> 
#> ℹ MOBSTER fits completed in 7.4s.
#> 
#> ── [ MOBSTER ] My MOBSTER model n = 200 with k = 2 Beta(s) without tail ────────
#> ● Clusters: π = 50% [C2] and 50% [C1], with π > 0.
#> ✖ No tail fit.
#> 
#> ● Beta C1 [n = 99, 50%] with mean = 0.6.
#> ● Beta C2 [n = 101, 50%] with mean = 0.12.
#> ℹ Score(s): NLL = -110.86; ICL = -188.81 (-188.81), H = 1.12 (1.12). Fit
#> converged by MM in 57 steps.
#>  [ MOBSTER fit ] 
#> 
#> ✔ Loaded input data, n = 200.
#> ❯ n = 200. Mixture with k = 1,2 Beta(s). Pareto tail: TRUE and FALSE. Output
#> clusters with π > 0.02 and n > 10.
#> ! mobster automatic setup FAST for the analysis.
#> ❯ Scoring (without parallel) 2 x 2 x 2 = 8 models by reICL.
#> 
#> [easypar] 2020-10-29 11:34:14 - Overriding parallel execution setup [FALSE] with global option : FALSE
#> 
#> 
#> ℹ MOBSTER fits completed in 9s.
#> 
#> ── [ MOBSTER ] My MOBSTER model n = 200 with k = 2 Beta(s) without tail ────────
#> ● Clusters: π = 53% [C1] and 47% [C2], with π > 0.
#> ✖ No tail fit.
#> 
#> ● Beta C1 [n = 106, 53%] with mean = 0.59.
#> ● Beta C2 [n = 94, 47%] with mean = 0.11.
#> ℹ Score(s): NLL = -117.56; ICL = -201.98 (-201.98), H = 1.35 (1.35). Fit
#> converged by MM in 76 steps.
#>  [ MOBSTER fit ] 
#> 
#> ✔ Loaded input data, n = 200.
#> ❯ n = 200. Mixture with k = 1,2 Beta(s). Pareto tail: TRUE and FALSE. Output
#> clusters with π > 0.02 and n > 10.
#> ! mobster automatic setup FAST for the analysis.
#> ❯ Scoring (without parallel) 2 x 2 x 2 = 8 models by reICL.
#> 
#> [easypar] 2020-10-29 11:34:23 - Overriding parallel execution setup [FALSE] with global option : FALSE
#> 
#> 
#> ℹ MOBSTER fits completed in 9.8s.
#> 
#> ── [ MOBSTER ] My MOBSTER model n = 200 with k = 2 Beta(s) without tail ────────
#> ● Clusters: π = 54% [C1] and 46% [C2], with π > 0.
#> ✖ No tail fit.
#> 
#> ● Beta C1 [n = 108, 54%] with mean = 0.59.
#> ● Beta C2 [n = 92, 46%] with mean = 0.13.
#> ℹ Score(s): NLL = -101.74; ICL = -170.66 (-170.66), H = 1.02 (1.02). Fit
#> converged by MM in 13 steps.
#>  [ MOBSTER fit ] 
#> 
#> ✔ Loaded input data, n = 200.
#> ❯ n = 200. Mixture with k = 1,2 Beta(s). Pareto tail: TRUE and FALSE. Output
#> clusters with π > 0.02 and n > 10.
#> ! mobster automatic setup FAST for the analysis.
#> ❯ Scoring (without parallel) 2 x 2 x 2 = 8 models by reICL.
#> 
#> [easypar] 2020-10-29 11:34:33 - Overriding parallel execution setup [FALSE] with global option : FALSE
#> 
#> 
#> ℹ MOBSTER fits completed in 9.9s.
#> 
#> ── [ MOBSTER ] My MOBSTER model n = 200 with k = 2 Beta(s) without tail ────────
#> ● Clusters: π = 58% [C1] and 42% [C2], with π > 0.
#> ✖ No tail fit.
#> 
#> ● Beta C1 [n = 114, 58%] with mean = 0.57.
#> ● Beta C2 [n = 86, 42%] with mean = 0.11.
#> ℹ Score(s): NLL = -105.37; ICL = -177.17 (-177.17), H = 1.78 (1.78). Fit
#> interrupted by MM in 100 steps.

boot_stats = bootstrapped_statistics(x$best, boot_results)

#> 
#> ── Computing model frequency ───────────────────────────────────────────────────
#> # A tibble: 1 x 3
#>   Model              Frequency fit.model
#>   <fct>                  <dbl> <lgl>    
#> 1 K = 2 without tail         1 TRUE     
#> 
#> ── Computing confidence Intervals (CI) for empirical quantiles ─────────────────
#> 
#> Mixing proportions 
#>  # A tibble: 3 x 8
#>   cluster statistics   min lower_quantile higher_quantile   max fit.value
#>   <chr>   <chr>      <dbl>          <dbl>           <dbl> <dbl>     <dbl>
#> 1 C1      Mixing pr… 0.499          0.500           0.574 0.578     0.517
#> 2 C2      Mixing pr… 0.422          0.426           0.500 0.501     0.483
#> 3 Tail    Mixing pr… 0              0               0     0         0    
#> # … with 1 more variable: init.value <dbl>
#> 
#> Tail shape/ scale 
#>  # A tibble: 0 x 8
#> # … with 8 variables: cluster <chr>, statistics <chr>, min <dbl>,
#> #   lower_quantile <dbl>, higher_quantile <dbl>, max <dbl>, fit.value <dbl>,
#> #   init.value <dbl>
#> 
#> Beta peaks 
#>  # A tibble: 4 x 8
#>   cluster statistics     min lower_quantile higher_quantile     max fit.value
#>   <chr>   <chr>        <dbl>          <dbl>           <dbl>   <dbl>     <dbl>
#> 1 C1      Mean       0.574          0.575           0.599   0.600     0.581  
#> 2 C1      Variance   0.00709        0.00712         0.00868 0.00874   0.00893
#> 3 C2      Mean       0.107          0.107           0.127   0.129     0.112  
#> 4 C2      Variance   0.00317        0.00317         0.00447 0.00455   0.00341
#> # … with 1 more variable: init.value <dbl>
#> 
#> ── Computing co-clustering probability for nonparametric bootstrap ─────────────
plot_bootstrap_mixing_proportions(x$best, boot_results, boot_stats)