Plot standardized beta profiles across continuous predictors

Fits a separate linear model for every predictor and displays the resulting standardized regression coefficients with 95% confidence intervals. This is the continuous-outcome companion to PlotZScore(): use it to compare the direction and magnitude of associations across a panel of predictors.

Usage

PlotBetaProfile(
  data,
  predictor_vars,
  outcome_var,
  covariates = NULL,
  VariableCategories = NULL,
  Sort = c("original", "pvalue", "fdr", "effect", "within_category_pvalue",
    "within_category_effect"),
  AdjustMethod = "fdr",
  Alpha = 0.05,
  Relabel = TRUE,
  codebook = NULL,
  RemoveXAxisLabels = TRUE,
  InteractiveLabels = TRUE
)

Arguments

data: A data frame containing the outcome, predictors, and covariates.
predictor_vars: Nonempty character vector of numeric predictor names.
outcome_var: Character string naming a numeric continuous outcome.
covariates: Optional character vector of covariate names. Covariates are included without automatic standardization.
VariableCategories: Optional categories for the predictors. Supply a vector corresponding to predictor_vars, a named vector keyed by predictor name, or a data frame with Variable and Category columns.
Sort: Variable ordering. One of "original", "pvalue", "fdr", "effect", "within_category_pvalue", or "within_category_effect".
AdjustMethod: Multiple-testing method passed to stats::p.adjust().
Alpha: Numeric significance threshold recorded in the returned metadata. Significance does not control plot color.
Relabel: Logical. If TRUE, display labels are resolved from the supplied codebook, then variable label attributes, with variable names as the fallback.
codebook: Optional data frame containing Variable and Label.
RemoveXAxisLabels: Logical. If TRUE, x-axis labels are hidden.
InteractiveLabels: Logical. If TRUE, the point layer contains a text aesthetic for plotly::ggplotly(..., tooltip = "text").

Value

A named list with three elements:

Plot: A ggplot object showing standardized betas and 95% CIs.
ResultsTable: A tibble with one successfully analyzed predictor per row and columns Variable, Label, Category, Beta, SE, CILow, CIHigh, PValue, FDR, N, R, AdjustedR, and Tooltip.
Metadata: A named list describing the outcome, requested and analyzed predictors, covariates, adjustment method, alpha threshold, sorting mode, confidence level, and number of fitted models.

Details

Each model has the form scale(outcome) ~ scale(predictor) + covariates. Covariates retain their original representation and are not standardized. The plotted beta is therefore the standard-deviation change in the outcome associated with a one-standard-deviation increase in the predictor, conditional on the supplied covariates. Complete cases are selected independently for every predictor model.

Examples

data(SampleData)
data(SampleVariableTypes)
Labelled <- RevalueData(SampleData, SampleVariableTypes)$RevaluedData
predictors <- c("Adiponectin", "C_Reactive_Protein", "Ferritin", "tau")

# Unadjusted
unadjusted <- PlotBetaProfile(
  data = Labelled,
  predictor_vars = predictors,
  outcome_var = "AXL"
)
unadjusted$Plot


# Adjusted
adjusted <- PlotBetaProfile(
  data = Labelled,
  predictor_vars = predictors,
  outcome_var = "AXL",
  covariates = c("age", "sex")
)
adjusted$ResultsTable
#> # A tibble: 4 × 13
#>   Variable   Label Category   Beta     SE   CILow CIHigh   PValue      FDR     N
#>   <chr>      <chr> <chr>     <dbl>  <dbl>   <dbl>  <dbl>    <dbl>    <dbl> <int>
#> 1 Adiponect… Adip… NA       0.0813 0.0556 -0.0281  0.191 1.45e- 1 1.93e- 1   322
#> 2 C_Reactiv… C-re… NA       0.0313 0.0558 -0.0785  0.141 5.75e- 1 5.75e- 1   322
#> 3 Ferritin   Ferr… NA       0.571  0.0466  0.480   0.663 1.31e-28 5.24e-28   322
#> 4 tau        Tau … NA       0.594  0.0542  0.487   0.701 1.48e-22 2.95e-22   223
#> # ℹ 3 more variables: R <dbl>, AdjustedR <dbl>, Tooltip <chr>

# Categories and within-category sorting
categories <- c("Metabolic", "Inflammation", "Inflammation", "Neurology")
categorized <- PlotBetaProfile(
  data = Labelled,
  predictor_vars = predictors,
  outcome_var = "AXL",
  covariates = c("age", "sex"),
  VariableCategories = categories,
  Sort = "within_category_pvalue"
)
categorized$Plot