jjstatsplot: Statistical Visualization for Clinical Research

Introduction

jjstatsplot provides comprehensive statistical visualization tools specifically designed for clinical research and pathology studies. Built on the powerful ggstatsplot package, it offers publication-ready plots with integrated statistical testing, making complex data analysis accessible to clinical researchers.

Learning Objectives:

Master statistical visualization for clinical data analysis
Learn publication-ready plot creation with integrated statistics
Apply appropriate statistical tests for different data types
Understand clinical interpretation of statistical visualizations
Implement quality control and reproducible reporting workflows

Module Overview

jjstatsplot encompasses four main areas of statistical visualization:

1. Categorical Data Analysis

Bar Charts: Group comparisons with chi-square testing
Pie Charts: Proportion visualization for categorical outcomes
Dot Plots: Summary statistics with confidence intervals
Mosaic Plots: Multi-dimensional categorical relationships

2. Continuous Data Analysis

Box-Violin Plots: Distribution comparisons between groups
Histograms: Distribution analysis with normality testing
Density Plots: Smooth distribution visualization
Within-Subject Comparisons: Paired data analysis

3. Correlation and Association

Scatter Plots: Relationship visualization with regression
Correlation Matrices: Multi-variable association analysis
Heatmaps: Pattern recognition in large datasets
Network Plots: Complex relationship visualization

4. Advanced Visualizations

Ridge Plots: Multiple distribution comparisons
Raincloud Plots: Combined distribution and individual points
Forest Plots: Meta-analysis and effect size visualization
Sankey Diagrams: Flow and pathway analysis

library(ClinicoPath)
library(dplyr)
library(ggplot2)

Getting Started

Clinical Data Examples

We’ll demonstrate jjstatsplot functions using clinical pathology datasets that represent real-world scenarios.

# Load clinical pathology datasets
data(histopathology)
data(breast_cancer_data)
data(ihc_molecular_comprehensive)

# Prepare clinical variables for analysis
clinical_data <- histopathology %>%
  mutate(
    # Convert variables to appropriate types
    Grade = factor(Grade, levels = 1:3, labels = c("Grade 1", "Grade 2", "Grade 3")),
    TStage = factor(TStage),
    Group = factor(Group),
    LVI = factor(LVI, levels = c("FALSE", "TRUE"), labels = c("Absent", "Present")),
    Age_Group = factor(ifelse(Age >= 65, "≥65 years", "<65 years")),
    
    # Create clinical outcome variables
    High_Risk = factor(ifelse(Grade == "Grade 3" | TStage %in% c("T3", "T4"), 
                             "High Risk", "Low Risk")),
    Recurrence_Risk = factor(case_when(
      Grade == "Grade 1" & TStage %in% c("T1", "T2") ~ "Low",
      Grade == "Grade 2" ~ "Intermediate", 
      Grade == "Grade 3" | TStage %in% c("T3", "T4") ~ "High",
      TRUE ~ "Intermediate"
    ), levels = c("Low", "Intermediate", "High"))
  )

cat("Clinical Dataset Overview:\n")
cat("Patients:", nrow(clinical_data), "\n")
cat("Variables:", ncol(clinical_data), "\n")

# Variable summary
clinical_summary <- clinical_data %>%
  summarise(
    median_age = median(Age, na.rm = TRUE),
    grade3_pct = round(mean(Grade == "Grade 3", na.rm = TRUE) * 100, 1),
    lvi_present_pct = round(mean(LVI == "Present", na.rm = TRUE) * 100, 1),
    high_risk_pct = round(mean(High_Risk == "High Risk", na.rm = TRUE) * 100, 1)
  )

print("Clinical Characteristics:")
print(clinical_summary)

Core Visualization Functions

1. Categorical Data Visualization

Bar Charts for Group Comparisons

Essential for comparing categorical outcomes between treatment groups.

# Bar Chart Example - Tumor Grade by Treatment Group
# In jamovi: JJStatsPlot > Categorical > Bar Charts

cat("Bar Chart Analysis - Tumor Grade by Treatment Group\n")
cat("==================================================\n\n")

# Basic bar chart with statistical testing
grade_by_group <- jjbarstats(
  data = clinical_data,
  dep = "Grade",              # Dependent variable (outcome)
  group = "Group",            # Grouping variable (treatment)
  title = "Tumor Grade Distribution by Treatment Group",
  subtitle = "Chi-square test for independence",
  xlab = "Treatment Group",
  ylab = "Count"
)

# Results interpretation
cat("Statistical Test: Chi-square test of independence\n")
cat("Null Hypothesis: Grade distribution is independent of treatment group\n")
cat("Alternative: Grade distribution differs between treatment groups\n\n")

# Clinical interpretation guidelines
cat("Clinical Interpretation Guidelines:\n")
cat("- Low-grade tumors (Grade 1-2): Better prognosis, less aggressive\n")
cat("- High-grade tumors (Grade 3): Poorer prognosis, more aggressive\n")
cat("- Treatment group differences may indicate selection bias or treatment effect\n")

Pie Charts for Proportion Visualization

Useful for showing the composition of categorical variables.

# Pie Chart Example - Recurrence Risk Distribution
# In jamovi: JJStatsPlot > Categorical > Pie Charts

cat("Pie Chart Analysis - Recurrence Risk Distribution\n")
cat("================================================\n\n")

recurrence_pie <- jjpiestats(
  data = clinical_data,
  dep = "Recurrence_Risk",
  title = "Patient Distribution by Recurrence Risk",
  subtitle = "Low, Intermediate, and High Risk Categories"
)

# Risk stratification summary
risk_distribution <- table(clinical_data$Recurrence_Risk)
risk_percentages <- round(prop.table(risk_distribution) * 100, 1)

cat("Recurrence Risk Distribution:\n")
for(i in 1:length(risk_distribution)) {
  cat(paste0(names(risk_distribution)[i], ": ", risk_distribution[i], 
            " patients (", risk_percentages[i], "%)\n"))
}

cat("\nClinical Utility:\n")
cat("- Risk stratification guides treatment intensity\n")
cat("- High-risk patients may benefit from adjuvant therapy\n")
cat("- Low-risk patients may be candidates for de-escalation\n")

2. Continuous Data Visualization

Box-Violin Plots for Group Comparisons

Standard approach for comparing continuous variables between groups.

# Box-Violin Plot Example - Age by Risk Group
# In jamovi: JJStatsPlot > Continuous > Between Groups

cat("Box-Violin Plot Analysis - Age by Risk Group\n")
cat("============================================\n\n")

age_by_risk <- jjbetweenstats(
  data = clinical_data,
  dep = "Age",                    # Continuous outcome variable
  group = "High_Risk",            # Grouping variable
  type = "parametric",            # Statistical test type
  pairwise.comparisons = TRUE,    # Post-hoc comparisons
  title = "Age Distribution by Risk Category",
  subtitle = "Two-sample t-test with effect size",
  xlab = "Risk Category",
  ylab = "Age (years)"
)

# Effect size interpretation
cat("Statistical Analysis:\n")
cat("Test: Two-sample t-test (parametric)\n")
cat("Effect size: Cohen's d\n")
cat("Assumptions: Normality, equal variances\n\n")

cat("Effect Size Interpretation (Cohen's d):\n")
cat("- Small effect: d = 0.2\n")
cat("- Medium effect: d = 0.5\n")
cat("- Large effect: d = 0.8\n\n")

cat("Clinical Relevance:\n")
cat("- Age may be a confounding factor in risk assessment\n")
cat("- Older patients may have different treatment tolerability\n")
cat("- Age-adjusted analysis may be necessary\n")

Histograms with Statistical Testing

Evaluate distribution characteristics and test against theoretical distributions.

# Histogram Example - Age Distribution Analysis
# In jamovi: JJStatsPlot > Continuous > Histograms

cat("Histogram Analysis - Age Distribution\n")
cat("====================================\n\n")

age_histogram <- jjhistostats(
  data = clinical_data,
  dep = "Age",
  test.value = 65,               # Test against clinically relevant age
  type = "parametric",
  normal.curve = TRUE,           # Overlay normal distribution
  title = "Patient Age Distribution",
  subtitle = "One-sample t-test against age 65",
  xlab = "Age (years)",
  binwidth = 5
)

# Distribution characteristics
age_stats <- clinical_data %>%
  summarise(
    n = sum(!is.na(Age)),
    mean_age = round(mean(Age, na.rm = TRUE), 1),
    median_age = median(Age, na.rm = TRUE),
    sd_age = round(sd(Age, na.rm = TRUE), 1),
    min_age = min(Age, na.rm = TRUE),
    max_age = max(Age, na.rm = TRUE),
    elderly_pct = round(mean(Age >= 65, na.rm = TRUE) * 100, 1)
  )

cat("Age Distribution Summary:\n")
cat("Sample size:", age_stats$n, "\n")
cat("Mean age:", age_stats$mean_age, "years\n")
cat("Median age:", age_stats$median_age, "years\n")
cat("Standard deviation:", age_stats$sd_age, "years\n")
cat("Age range:", age_stats$min_age, "-", age_stats$max_age, "years\n")
cat("Elderly patients (≥65):", age_stats$elderly_pct, "%\n\n")

cat("Clinical Implications:\n")
cat("- Patient age distribution affects generalizability\n")
cat("- Elderly patients may require modified treatment protocols\n")
cat("- Age-related comorbidities impact treatment selection\n")

3. Correlation and Association Analysis

Scatter Plots with Regression Analysis

Examine relationships between continuous variables.

# Scatter Plot Example - Biomarker Correlation
# In jamovi: JJStatsPlot > Correlation > Scatter Plots

cat("Scatter Plot Analysis - Biomarker Correlation\n")
cat("=============================================\n\n")

# Using measurement variables from histopathology data
biomarker_scatter <- jjscatterstats(
  data = clinical_data,
  x = "MeasurementA",           # First biomarker
  y = "MeasurementB",           # Second biomarker
  type = "parametric",
  conf.level = 0.95,
  title = "Biomarker A vs Biomarker B Correlation",
  subtitle = "Pearson correlation with 95% confidence interval",
  xlab = "Biomarker A Expression",
  ylab = "Biomarker B Expression"
)

# Correlation strength interpretation
cat("Correlation Strength Interpretation:\n")
cat("- Negligible: |r| < 0.1\n")
cat("- Weak: 0.1 ≤ |r| < 0.3\n")
cat("- Moderate: 0.3 ≤ |r| < 0.5\n")
cat("- Strong: 0.5 ≤ |r| < 0.7\n")
cat("- Very strong: |r| ≥ 0.7\n\n")

cat("Clinical Applications:\n")
cat("- Biomarker co-expression patterns\n")
cat("- Pathway analysis and validation\n")
cat("- Quality control for technical replicates\n")
cat("- Surrogate endpoint validation\n")

Correlation Matrix for Multiple Variables

Comprehensive analysis of relationships between multiple variables.

# Correlation Matrix Example - Clinical Variables
# In jamovi: JJStatsPlot > Correlation > Correlation Matrix

cat("Correlation Matrix Analysis - Clinical Variables\n")
cat("===============================================\n\n")

# Select continuous clinical variables
continuous_vars <- clinical_data %>%
  select(Age, MeasurementA, MeasurementB, OverallTime) %>%
  na.omit()

if(nrow(continuous_vars) > 10) {
  correlation_matrix <- jjcorrmat(
    data = continuous_vars,
    type = "parametric",
    title = "Clinical Variables Correlation Matrix",
    subtitle = "Pearson correlations with significance testing"
  )
  
  cat("Matrix Interpretation:\n")
  cat("- Diagonal elements are always 1.0 (self-correlation)\n")
  cat("- Upper triangle shows correlation coefficients\n")
  cat("- Lower triangle may show significance levels\n")
  cat("- Color intensity indicates correlation strength\n\n")
  
  cat("Clinical Value:\n")
  cat("- Identify multicollinearity in regression models\n")
  cat("- Discover unexpected relationships between variables\n")
  cat("- Guide feature selection for predictive models\n")
  cat("- Quality assurance for related measurements\n")
}

Advanced Clinical Applications

Paired Data Analysis

Essential for before-after comparisons and matched samples.

# Paired Analysis Example - Pre/Post Treatment Comparison
# In jamovi: JJStatsPlot > Continuous > Within Subject

cat("Paired Data Analysis - Treatment Response\n")
cat("========================================\n\n")

# Create synthetic paired data for demonstration
set.seed(123)
paired_data <- clinical_data %>%
  slice_head(n = 50) %>%
  mutate(
    Pre_Treatment = MeasurementA,
    Post_Treatment = MeasurementA + rnorm(50, mean = -2, sd = 1.5),
    Patient_ID = paste0("PT", sprintf("%03d", 1:50))
  ) %>%
  select(Patient_ID, Pre_Treatment, Post_Treatment) %>%
  tidyr::pivot_longer(cols = c(Pre_Treatment, Post_Treatment),
                     names_to = "Timepoint", 
                     values_to = "Biomarker_Level")

if(nrow(paired_data) > 0) {
  paired_comparison <- jjwithinstats(
    data = paired_data,
    x = "Timepoint",
    y = "Biomarker_Level",
    id = "Patient_ID",
    paired = TRUE,
    type = "parametric",
    title = "Pre vs Post-Treatment Biomarker Levels",
    subtitle = "Paired t-test with individual trajectories",
    xlab = "Treatment Timepoint",
    ylab = "Biomarker Level"
  )
  
  cat("Paired Analysis Advantages:\n")
  cat("- Controls for individual patient baseline differences\n")
  cat("- Higher statistical power than independent samples\n")
  cat("- Shows individual patient trajectories\n")
  cat("- Appropriate for longitudinal studies\n\n")
  
  cat("Clinical Applications:\n")
  cat("- Treatment response monitoring\n")
  cat("- Before-after intervention studies\n")
  cat("- Matched case-control designs\n")
  cat("- Quality of life assessments\n")
}

Subgroup Analysis

Evaluate treatment effects across different patient populations.

cat("Subgroup Analysis - Treatment Effect by Grade\n")
cat("=============================================\n\n")

# Subgroup analysis by tumor grade
subgroup_results <- list()

for(grade_level in c("Grade 1", "Grade 2", "Grade 3")) {
  subgroup_data <- clinical_data %>%
    filter(Grade == grade_level & !is.na(MeasurementA))
  
  if(nrow(subgroup_data) >= 10) {
    cat(paste0("Analysis for ", grade_level, ":\n"))
    cat(paste0("Sample size: ", nrow(subgroup_data), "\n"))
    
    # Age comparison by group within grade
    age_summary <- subgroup_data %>%
      group_by(Group) %>%
      summarise(
        n = n(),
        mean_age = round(mean(Age, na.rm = TRUE), 1),
        sd_age = round(sd(Age, na.rm = TRUE), 1),
        .groups = 'drop'
      )
    
    print(age_summary)
    cat("\n")
  }
}

cat("Subgroup Analysis Considerations:\n")
cat("- Reduced sample sizes may limit statistical power\n")
cat("- Multiple comparisons increase Type I error risk\n")
cat("- Pre-planned vs post-hoc analyses have different interpretations\n")
cat("- Clinical significance may differ from statistical significance\n")

Integration with Clinical Workflows

Quality Control Applications

Statistical visualization for laboratory and clinical QC.

cat("Quality Control Applications\n")
cat("===========================\n\n")

# QC metrics analysis
qc_data <- clinical_data %>%
  mutate(
    # Simulate QC metrics
    Processing_Batch = sample(1:5, n(), replace = TRUE),
    Staining_Quality = sample(c("Excellent", "Good", "Acceptable", "Poor"), 
                             n(), replace = TRUE, prob = c(0.4, 0.3, 0.2, 0.1)),
    Measurement_CV = abs(rnorm(n(), mean = 5, sd = 2))  # Coefficient of variation
  ) %>%
  filter(!is.na(MeasurementA))

# Batch effect analysis
if(nrow(qc_data) > 20) {
  cat("Batch Effect Analysis:\n")
  batch_summary <- qc_data %>%
    group_by(Processing_Batch) %>%
    summarise(
      n = n(),
      mean_measurement = round(mean(MeasurementA, na.rm = TRUE), 2),
      sd_measurement = round(sd(MeasurementA, na.rm = TRUE), 2),
      cv_measurement = round(sd_measurement / mean_measurement * 100, 1),
      .groups = 'drop'
    )
  
  print(batch_summary)
  
  cat("\nQuality Control Guidelines:\n")
  cat("- CV <10%: Excellent reproducibility\n")
  cat("- CV 10-15%: Good reproducibility\n") 
  cat("- CV 15-20%: Acceptable reproducibility\n")
  cat("- CV >20%: Poor reproducibility, investigate\n\n")
}

cat("QC Visualization Applications:\n")
cat("- Monitor analytical performance over time\n")
cat("- Detect systematic errors and drift\n")
cat("- Validate new analytical methods\n")
cat("- Support accreditation requirements\n")

Best Practices for Clinical Research

Statistical Considerations

cat("Statistical Best Practices for Clinical Visualization\n")
cat("====================================================\n\n")

cat("1. Choose Appropriate Statistical Tests:\n")
cat("   - Parametric: Normal distribution, continuous data\n")
cat("   - Non-parametric: Skewed data, ordinal scales\n")
cat("   - Robust: Outliers present, mixed distributions\n")
cat("   - Bayesian: Small samples, prior information available\n\n")

cat("2. Effect Size Reporting:\n")
cat("   - Always report effect sizes alongside p-values\n")
cat("   - Consider clinical significance, not just statistical significance\n")
cat("   - Use confidence intervals for uncertainty quantification\n\n")

cat("3. Multiple Comparisons:\n")
cat("   - Adjust p-values when testing multiple hypotheses\n")
cat("   - Pre-specify primary and secondary endpoints\n")
cat("   - Consider false discovery rate control\n\n")

cat("4. Sample Size Considerations:\n")
cat("   - Ensure adequate power for primary analyses\n")
cat("   - Be cautious with subgroup analyses (small n)\n")
cat("   - Report confidence intervals to show precision\n\n")

cat("5. Missing Data:\n")
cat("   - Report patterns and extent of missing data\n")
cat("   - Use appropriate missing data methods\n")
cat("   - Conduct sensitivity analyses\n")

Publication Guidelines

cat("Publication-Ready Visualization Guidelines\n")
cat("==========================================\n\n")

cat("1. Figure Quality:\n")
cat("   - High resolution (≥300 DPI) for print\n")
cat("   - Clear, readable fonts (≥10 point)\n")
cat("   - Appropriate color schemes (colorblind-friendly)\n")
cat("   - Consistent styling across figures\n\n")

cat("2. Statistical Reporting:\n")
cat("   - Include sample sizes in figure legends\n")
cat("   - Report exact p-values when possible\n")
cat("   - Specify statistical tests used\n")
cat("   - Include effect sizes and confidence intervals\n\n")

cat("3. Clinical Context:\n")
cat("   - Provide clinical interpretation of findings\n")
cat("   - Discuss clinical significance vs statistical significance\n")
cat("   - Include relevant reference ranges or cutpoints\n")
cat("   - Consider real-world application\n\n")

cat("4. Reproducibility:\n")
cat("   - Document software versions\n")
cat("   - Provide clear methods description\n")
cat("   - Share analysis code when possible\n")
cat("   - Include raw data summaries\n")

Conclusion

jjstatsplot provides comprehensive statistical visualization tools essential for modern clinical research and pathology studies. The integration of statistical testing with publication-ready graphics makes complex data analysis accessible to clinical researchers while maintaining scientific rigor.

Key Advantages

Integrated Statistics: Automatic statistical testing with visualization
Clinical Focus: Designed for healthcare research applications
Publication Ready: High-quality graphics suitable for journals
User Friendly: Accessible through jamovi interface
Comprehensive: Covers all major visualization needs

Next Steps

To explore specific visualization techniques:

Categorical Analysis: See detailed categorical plot vignettes
Continuous Data: Explore advanced continuous data visualization
Correlation Studies: Learn comprehensive correlation analysis
Clinical Applications: Study real-world clinical research examples

This introduction provides the foundation for using jjstatsplot in clinical research. Explore the detailed vignettes for specific techniques and advanced applications.

ClinicoPath Development Team

2025-06-30