Skip to contents

Univariate Tests for Multi-Level Categorical Predictors

Source: R/cyt_univariate_multi.R
cyt_univariate_multi.Rd

cyt_univariate_multi provides univariate statistical testing for categorical predictors with more than two levels. For one-way designs, each categorical predictor and numeric outcome pair receives a global test followed by pairwise comparisons when appropriate. Users may choose between classical ANOVA with Tukey's Honest Significant Difference (HSD) or a non-parametric Kruskal-Wallis test followed by pairwise Wilcoxon rank-sum tests. The function also supports explicit two-way ANOVA and ANCOVA designs with optional primary:secondary and primary:covariate interaction terms.

Usage

cyt_univariate_multi(
  data,
  method = c("anova", "kruskal"),
  design = c("one_way", "two_way", "ancova"),
  cat_vars = NULL,
  cont_vars = NULL,
  p_adjust_method = "BH",
  primary_cat_var = NULL,
  secondary_cat_var = NULL,
  covariate_col = NULL,
  include_primary_secondary_interaction = FALSE,
  include_primary_covariate_interaction = FALSE,
  format_output = FALSE,
  progress = NULL
)

Arguments

data

A data frame or matrix containing both categorical and continuous variables. Character columns will be converted to factors.

method

Character specifying the type of global test to perform. Use "anova" (default) for one-way ANOVA with Tukey HSD or "kruskal" for Kruskal-Wallis with pairwise Wilcoxon tests. "kruskal" is only supported when design = "one_way".

design

Character specifying the model design. Use "one_way" (default) to preserve the existing multi-level ANOVA/Kruskal-Wallis workflow, "two_way" for two-way ANOVA, or "ancova" for ANCOVA.

cat_vars

Optional character vector of predictor column names used by the one-way workflow. When NULL, all factor or character columns in data are used.

cont_vars

Optional character vector of numeric outcome variable names. When NULL, all numeric columns in data are used, excluding covariate_col for ANCOVA.

p_adjust_method

Character string specifying the method for p-value adjustment across pairwise Kruskal-Wallis follow-up comparisons. Passed to adjust_p. Default is "BH".

primary_cat_var

Optional primary categorical predictor used by the two-way ANOVA and ANCOVA workflows.

secondary_cat_var

Optional secondary categorical predictor used by the two-way ANOVA and ANCOVA workflows.

covariate_col

Optional numeric covariate used by the ANCOVA workflow.

include_primary_secondary_interaction

Logical. Whether to include the primary:secondary interaction term when secondary_cat_var is supplied.

include_primary_covariate_interaction

Logical. Whether to include the primary:covariate interaction term when covariate_col is supplied.

format_output

Logical. If TRUE, returns a list with global results, pairwise results, and assumption summaries; otherwise (default) returns a list of numeric vectors keyed by "Outcome_Categorical" for the one-way workflow. Two-way ANOVA and ANCOVA return raw model bundles keyed by outcome when format_output = FALSE.

progress

Optional. A Shiny Progress object for reporting progress updates during the analysis.

Value

Either a list of adjusted pairwise p-values (if format_output = FALSE and design = "one_way") or a list with results, pairwise, and assumptions data frames (if format_output = TRUE).

Author

Shubh Saraswat

Examples

data("ExampleData1")
cyt_univariate_multi(ExampleData1[, c(1:2, 5:6)], method = "kruskal",
  format_output = TRUE
)
#> $results
#>   Outcome Categorical Comparison           Test Statistic P_value
#> 1  GM.CSF       Group    Overall Kruskal-Wallis     0.300   0.861
#> 2   IFN.G       Group    Overall Kruskal-Wallis     8.127   0.017
#> 3  GM.CSF   Treatment    Overall Kruskal-Wallis   198.241   0.000
#> 4   IFN.G   Treatment    Overall Kruskal-Wallis   191.334   0.000
#> 
#> $pairwise
#>                    Outcome Categorical               Comparison
#> IFN.G_Group.1        IFN.G       Group             ND vs PreT2D
#> IFN.G_Group.2        IFN.G       Group                ND vs T2D
#> IFN.G_Group.3        IFN.G       Group            PreT2D vs T2D
#> GM.CSF_Treatment.1  GM.CSF   Treatment          CD3/CD28 vs LPS
#> GM.CSF_Treatment.2  GM.CSF   Treatment CD3/CD28 vs Unstimulated
#> GM.CSF_Treatment.3  GM.CSF   Treatment      LPS vs Unstimulated
#> IFN.G_Treatment.1    IFN.G   Treatment          CD3/CD28 vs LPS
#> IFN.G_Treatment.2    IFN.G   Treatment CD3/CD28 vs Unstimulated
#> IFN.G_Treatment.3    IFN.G   Treatment      LPS vs Unstimulated
#>                                 Test Estimate P_adj
#> IFN.G_Group.1      Pairwise Wilcoxon    36.35 0.025
#> IFN.G_Group.2      Pairwise Wilcoxon   -16.63 0.552
#> IFN.G_Group.3      Pairwise Wilcoxon   -65.43 0.025
#> GM.CSF_Treatment.1 Pairwise Wilcoxon     3.73 0.000
#> GM.CSF_Treatment.2 Pairwise Wilcoxon     4.53 0.000
#> GM.CSF_Treatment.3 Pairwise Wilcoxon     0.59 0.000
#> IFN.G_Treatment.1  Pairwise Wilcoxon 45492.85 0.000
#> IFN.G_Treatment.2  Pairwise Wilcoxon 46319.12 0.000
#> IFN.G_Treatment.3  Pairwise Wilcoxon    77.53 0.000
#> 
#> $assumptions
#> NULL
#>