Comprehensive Guide to Enhanced Statistical Visualization with jggstats

Enhanced Statistical Visualization with jggstats

Overview

The jggstats function provides advanced statistical visualization capabilities using the ggstats package. It offers multiple analysis types for different statistical contexts:

Model Coefficients: Visualize regression coefficients with confidence intervals
Model Comparisons: Compare multiple models side-by-side
Likert Plots: Specialized visualization for Likert scale data
Survey Analysis: Weighted survey data visualization
Proportion Analysis: Categorical data proportions
Cross-tabulation: Relationship between categorical variables
Weighted Means: Group comparisons with survey weights
Cascade Analysis: Data filtering and sample size tracking

Understanding Statistical Visualization

When to Use Each Analysis Type

Model Coefficients are ideal for: - Regression model interpretation - Effect size visualization - Confidence interval display - Variable importance assessment

Model Comparisons excel at: - Comparing nested models - Model selection visualization - Effect stability across models - Sensitivity analysis display

Likert Plots are perfect for: - Survey response visualization - Attitude scale analysis - Agreement/disagreement patterns - Ordered categorical data

Survey Analysis handles: - Complex survey designs - Weighted population estimates - Demographic breakdowns - Representative sampling

Basic Usage Examples

Model Coefficient Visualization

# Load test data
data("linear_model_data")

# Basic coefficient plot
# jggstats(
#   data = linear_model_data,
#   dependent_var = "outcome",
#   independent_vars = c("age", "sex", "treatment"),
#   analysis_type = "ggcoef_model",
#   model_type = "lm",
#   plot_title = "Linear Model Coefficients"
# )

Logistic Regression Coefficients

# Load logistic model data
data("logistic_model_data")

# Logistic regression coefficient plot
# jggstats(
#   data = logistic_model_data,
#   dependent_var = "disease",
#   independent_vars = c("age", "sex", "smoking", "bmi"),
#   analysis_type = "ggcoef_model",
#   model_type = "glm",
#   family = "binomial",
#   sort_coefficients = TRUE,
#   show_significance = TRUE,
#   plot_title = "Disease Risk Factors",
#   plot_subtitle = "Logistic regression odds ratios"
# )

Clinical Research Applications

Survival Analysis Visualization

# Load survival data
data("survival_analysis_data")

# Cox regression coefficient plot
# jggstats(
#   data = survival_analysis_data,
#   model_formula = "Surv(survival_time, event) ~ age + sex + stage + treatment",
#   analysis_type = "ggcoef_model",
#   model_type = "coxph",
#   show_intercept = FALSE,
#   sort_coefficients = TRUE,
#   color_palette = "viridis",
#   plot_title = "Survival Analysis",
#   plot_subtitle = "Cox regression hazard ratios",
#   show_model_summary = TRUE,
#   show_interpretation = TRUE
# )

Biomarker Analysis

# Focus on biomarker effects
biomarker_data <- logistic_model_data %>%
  filter(!is.na(biomarker_level))

# jggstats(
#   data = biomarker_data,
#   dependent_var = "disease",
#   independent_vars = c("biomarker_level", "age", "sex"),
#   grouping_var = "family_history",
#   analysis_type = "ggcoef_compare",
#   model_type = "glm",
#   family = "binomial",
#   confidence_level = 0.95,
#   color_palette = "set1",
#   plot_title = "Biomarker Effects by Family History",
#   facet_var = "clinic",
#   facet_type = "wrap"
# )

Treatment Response Analysis

# Analyze treatment effects
# jggstats(
#   data = linear_model_data,
#   dependent_var = "outcome",
#   independent_vars = c("treatment", "age", "baseline_score"),
#   analysis_type = "ggcoef_model",
#   model_type = "lm",
#   show_intercept = FALSE,
#   standardized = TRUE,
#   show_significance = TRUE,
#   plot_title = "Treatment Effects Analysis",
#   plot_subtitle = "Standardized coefficients",
#   theme_style = "classic"
# )

Survey Data Analysis

Likert Scale Visualization

# Load survey data
data("likert_survey_data")

# Likert plot for satisfaction measures
# jggstats(
#   data = likert_survey_data,
#   dependent_var = "satisfaction_work",
#   grouping_var = "department",
#   analysis_type = "gglikert",
#   likert_levels = 5,
#   color_palette = "viridis",
#   plot_title = "Work Satisfaction by Department",
#   plot_subtitle = "Employee survey responses (1-5 scale)"
# )

Weighted Survey Analysis

# Load survey proportion data
data("survey_proportion_data")

# Weighted survey analysis
# jggstats(
#   data = survey_proportion_data,
#   dependent_var = "political_preference",
#   weight_var = "weight",
#   grouping_var = "age_category",
#   analysis_type = "ggsurvey",
#   color_palette = "dark2",
#   plot_title = "Political Preferences by Age",
#   plot_subtitle = "Weighted survey estimates",
#   show_interpretation = TRUE
# )

Cross-tabulation Analysis

# Load cross-tabulation data
data("cross_tab_data")

# Cross-tabulation visualization
# jggstats(
#   data = cross_tab_data,
#   dependent_var = "test_result",
#   grouping_var = "disease_status",
#   analysis_type = "stat_cross",
#   color_palette = "paired",
#   plot_title = "Diagnostic Test Performance",
#   plot_subtitle = "Test results by true disease status",
#   facet_var = "age_group",
#   facet_type = "wrap"
# )

Educational Data Analysis

Academic Achievement Modeling

# Load educational data
data("model_comparison_data")

# Model achievement predictors
# jggstats(
#   data = model_comparison_data,
#   dependent_var = "test_score",
#   independent_vars = c("socioeconomic_status", "parent_education", "school_type", "class_size"),
#   analysis_type = "ggcoef_model",
#   model_type = "lm",
#   sort_coefficients = TRUE,
#   show_significance = TRUE,
#   plot_title = "Academic Achievement Predictors",
#   plot_subtitle = "Factors affecting test scores",
#   show_model_summary = TRUE
# )

School Comparison Analysis

# Compare models across districts
# jggstats(
#   data = model_comparison_data,
#   dependent_var = "test_score",
#   independent_vars = c("socioeconomic_status", "parent_education"),
#   grouping_var = "district",
#   analysis_type = "ggcoef_compare",
#   model_type = "lm",
#   confidence_level = 0.90,
#   color_palette = "set1",
#   plot_title = "District-Specific Achievement Models",
#   plot_subtitle = "Model comparison across school districts"
# )

Advanced Statistical Modeling

Mixed Effects Models

# Load longitudinal data
data("mixed_effects_data")

# Mixed effects model visualization
# jggstats(
#   data = mixed_effects_data,
#   dependent_var = "score",
#   independent_vars = c("time_months", "treatment_group", "baseline_severity"),
#   analysis_type = "ggcoef_model",
#   model_type = "lmer",
#   show_intercept = FALSE,
#   sort_coefficients = TRUE,
#   plot_title = "Longitudinal Treatment Effects",
#   plot_subtitle = "Mixed effects model coefficients",
#   theme_style = "minimal"
# )

Complex Survey Design

# Load complex survey data
data("weighted_analysis_data")

# Weighted means analysis
# jggstats(
#   data = weighted_analysis_data,
#   dependent_var = "income_thousands",
#   grouping_var = "head_education",
#   weight_var = "final_weight",
#   analysis_type = "stat_weighted_mean",
#   color_palette = "viridis",
#   plot_title = "Income by Education Level",
#   plot_subtitle = "Population-weighted estimates",
#   show_interpretation = TRUE,
#   output_format = "both"
# )

Proportion Analysis

# Analyze health coverage proportions
# jggstats(
#   data = weighted_analysis_data,
#   dependent_var = "insurance_coverage",
#   grouping_var = "geography",
#   analysis_type = "stat_prop",
#   color_palette = "dark2",
#   plot_title = "Insurance Coverage by Geography",
#   plot_subtitle = "Proportion analysis",
#   facet_var = "state_region",
#   facet_type = "wrap"
# )

Data Quality and Cascade Analysis

Sample Size Tracking

# Cascade analysis for data quality
# jggstats(
#   data = linear_model_data,
#   dependent_var = "outcome",
#   independent_vars = c("age", "sex", "treatment", "baseline_score"),
#   analysis_type = "ggcascade",
#   plot_title = "Data Cascade Analysis",
#   plot_subtitle = "Sample size at each analysis step",
#   theme_style = "classic"
# )

Missing Data Patterns

# Create data with missing values
incomplete_data <- model_comparison_data
incomplete_data$test_score[sample(nrow(incomplete_data), 50)] <- NA
incomplete_data$socioeconomic_status[sample(nrow(incomplete_data), 30)] <- NA

# jggstats(
#   data = incomplete_data,
#   dependent_var = "test_score",
#   independent_vars = c("socioeconomic_status", "parent_education", "school_type"),
#   analysis_type = "ggcascade",
#   plot_title = "Missing Data Impact",
#   plot_subtitle = "Sample reduction due to missing values"
# )

Advanced Customization

Custom Styling and Themes

# Advanced styling example
# jggstats(
#   data = logistic_model_data,
#   dependent_var = "disease",
#   independent_vars = c("age", "bmi", "biomarker_level"),
#   grouping_var = "smoking",
#   analysis_type = "ggcoef_model",
#   model_type = "glm",
#   family = "binomial",
#   standardized = FALSE,
#   show_intercept = FALSE,
#   sort_coefficients = TRUE,
#   show_significance = TRUE,
#   confidence_level = 0.95,
#   color_palette = "viridis",
#   theme_style = "dark",
#   plot_title = "Disease Risk Model",
#   plot_subtitle = "Logistic regression with custom styling",
#   x_label = "Log Odds Ratio",
#   y_label = "Risk Factors",
#   show_model_summary = TRUE,
#   show_interpretation = TRUE
# )

Faceted Analysis

# Multi-panel analysis
# jggstats(
#   data = survey_proportion_data,
#   dependent_var = "approve_policy",
#   grouping_var = "political_preference",
#   analysis_type = "stat_prop",
#   facet_var = "education_level",
#   facet_type = "wrap",
#   color_palette = "set1",
#   plot_title = "Policy Approval by Political Preference",
#   plot_subtitle = "Broken down by education level",
#   theme_style = "minimal"
# )

Model Comparison and Selection

Comparing Different Model Types

# Compare different approaches to the same data
base_data <- linear_model_data

# Model 1: Simple linear
# jggstats(
#   data = base_data,
#   dependent_var = "outcome",
#   independent_vars = c("age", "sex"),
#   analysis_type = "ggcoef_model",
#   model_type = "lm",
#   plot_title = "Simple Linear Model",
#   show_model_summary = TRUE
# )

# Model 2: With interactions
# jggstats(
#   data = base_data,
#   model_formula = "outcome ~ age * sex + treatment",
#   analysis_type = "ggcoef_model",
#   model_type = "lm",
#   plot_title = "Model with Interactions",
#   show_model_summary = TRUE
# )

Nested Model Comparison

# Systematic model comparison
# jggstats(
#   data = model_comparison_data,
#   dependent_var = "test_score",
#   independent_vars = c("socioeconomic_status", "parent_education"),
#   grouping_var = "school_type",
#   analysis_type = "ggcoef_compare",
#   model_type = "lm",
#   sort_coefficients = TRUE,
#   plot_title = "Nested Model Comparison",
#   plot_subtitle = "Effects across different school types",
#   output_format = "both"
# )

Interpretation Guidelines

Understanding Coefficient Plots

Key elements to interpret:

Point estimates: The central value of the effect
Confidence intervals: Uncertainty around the estimate
Significance: Whether intervals cross zero (for coefficients) or one (for odds ratios)
Magnitude: The practical importance of effects
Direction: Positive or negative associations

Statistical Significance

# Example with significance markers
# jggstats(
#   data = logistic_model_data,
#   dependent_var = "disease",
#   independent_vars = c("age", "sex", "smoking", "family_history", "bmi"),
#   analysis_type = "ggcoef_model",
#   model_type = "glm",
#   family = "binomial",
#   show_significance = TRUE,
#   sort_coefficients = TRUE,
#   plot_title = "Significance Testing Example",
#   plot_subtitle = "Stars indicate statistical significance",
#   show_interpretation = TRUE
# )

Performance Considerations

Large Dataset Handling

For large datasets (>10,000 observations), consider:

Model complexity: Simpler models fit faster
Variable selection: Focus on key predictors
Sampling: Use representative subsets for exploration

# Example with larger dataset simulation
large_data <- do.call(rbind, replicate(5, linear_model_data, simplify = FALSE))
large_data$id <- 1:nrow(large_data)

# Efficient analysis
# jggstats(
#   data = large_data,
#   dependent_var = "outcome",
#   independent_vars = c("age", "sex", "treatment"),
#   analysis_type = "ggcoef_model",
#   model_type = "lm",
#   plot_title = "Large Dataset Analysis",
#   plot_subtitle = paste("N =", nrow(large_data), "observations")
# )

Troubleshooting Guide

Common Issues and Solutions

Issue: “Model convergence problems” - Solution: Check for perfect separation in logistic models - Solution: Consider variable scaling for continuous predictors - Solution: Remove highly correlated predictors

Issue: “Confidence intervals too wide” - Solution: Increase sample size if possible - Solution: Consider variable transformation - Solution: Check for outliers affecting estimates

Issue: “Plot readability problems” - Solution: Reduce number of variables displayed - Solution: Use faceting for complex comparisons - Solution: Adjust text size and plot dimensions

Data Quality Checks

# Function to check data suitability
check_modeling_data <- function(data, dependent_var, independent_vars) {
  
  cat("Data Quality Assessment for Statistical Modeling\n")
  cat("=============================================\n\n")
  
  # Basic info
  cat("Dataset dimensions:", nrow(data), "rows,", ncol(data), "columns\n\n")
  
  # Check dependent variable
  if (!is.null(dependent_var)) {
    dep_data <- data[[dependent_var]]
    cat("Dependent Variable (", dependent_var, "):\n")
    cat("  - Type:", class(dep_data)[1], "\n")
    cat("  - Missing values:", sum(is.na(dep_data)), "/", length(dep_data), "\n")
    if (is.numeric(dep_data)) {
      cat("  - Range:", round(min(dep_data, na.rm = TRUE), 2), "to", 
          round(max(dep_data, na.rm = TRUE), 2), "\n")
    } else {
      cat("  - Unique values:", length(unique(dep_data[!is.na(dep_data)])), "\n")
    }
    cat("\n")
  }
  
  # Check independent variables
  if (!is.null(independent_vars)) {
    cat("Independent Variables:\n")
    for (var in independent_vars) {
      if (var %in% names(data)) {
        var_data <- data[[var]]
        cat("  -", var, ":")
        cat(" Type:", class(var_data)[1])
        cat(", Missing:", sum(is.na(var_data)))
        if (is.numeric(var_data)) {
          cat(", Range:", round(min(var_data, na.rm = TRUE), 2), "-", 
              round(max(var_data, na.rm = TRUE), 2))
        } else {
          cat(", Levels:", length(unique(var_data[!is.na(var_data)])))
        }
        cat("\n")
      }
    }
  }
}

# Example usage
check_modeling_data(
  data = linear_model_data,
  dependent_var = "outcome",
  independent_vars = c("age", "sex", "treatment")
)

Best Practices

Analysis Workflow

Exploratory Analysis: Start with descriptive statistics
Model Building: Begin with simple models, add complexity gradually
Diagnostics: Check model assumptions and fit
Visualization: Use coefficient plots for interpretation
Validation: Cross-validate findings when possible

Visualization Principles

Clarity: Keep plots simple and focused
Consistency: Use consistent color schemes and styling
Context: Provide adequate titles and labels
Accuracy: Ensure statistical correctness
Accessibility: Consider colorblind-friendly palettes

Conclusion

The jggstats function provides comprehensive tools for statistical visualization across multiple domains. Whether analyzing clinical trials, survey data, or educational outcomes, these enhanced visualizations help communicate statistical findings effectively.

Key advantages:

Comprehensive: Supports multiple analysis types
Flexible: Extensive customization options
Accurate: Built on robust statistical foundations
Interpretable: Clear visual communication of results
Scalable: Handles various dataset sizes and complexities

For additional examples and advanced techniques, explore the comprehensive test datasets included with the package.

ClinicoPath Package

2025-07-13