Introduction to Saqrmisc: Comprehensive Data Analysis and Visualization

Introduction

The Saqrmisc package provides a comprehensive suite of functions for data processing, statistical analysis, and visualization. This vignette demonstrates the main features and capabilities of the package through practical examples.

Installation and Loading

# Install from GitHub (if not already installed)
# devtools::install_github("mohsaqr/Saqrmisc")

# Load the package
library(Saqrmisc)

Package Overview

The Saqrmisc package includes ten main functional areas:

Group Comparisons (compare_groups)
Correlation Matrix (correlation_matrix)
Full Correlation Tables (correlations)
Descriptive Statistics (descriptive_table)
Categorical Frequency Tables (categorical_table)
Data Transformation (center, scale_vars, standardize, reverse_code)
Missing Data Analysis (missing_analysis, replace_missing)
Model-Based Clustering (clustering)
Categorical Variable Analysis (mosaic_analysis)
Network Estimation (estimate_single_network)

API Design

All functions use a consistent quoted-string API for variable names:

# Variable names as quoted strings
compare_groups(data, category = "gender", Vars = c("score1", "score2"))
correlations(data, Vars = c("x", "y", "z"), group_by = "group")
center(data, Vars = "score", group_by = "cluster")

# NULL for auto-selection of all numeric variables
correlations(data)  # Correlates all numeric variables
correlation_matrix(data)  # Matrix of all numeric variables

1. Group Comparisons

The compare_groups() function creates comprehensive comparison plots using ggbetweenstats to compare groups across multiple variables.

Example: Treatment Effect Analysis

# Generate sample treatment data
set.seed(456)
treatment_data <- data.frame(
  patient_id = 1:200,
  group = sample(c("Control", "Treatment"), 200, replace = TRUE),
  anxiety = rnorm(200, mean = 5, sd = 1.5),
  depression = rnorm(200, mean = 4, sd = 1.2),
  stress = rnorm(200, mean = 6, sd = 1.8),
  study_site = sample(c("Site_A", "Site_B", "Site_C"), 200, replace = TRUE),
  gender = sample(c("Male", "Female"), 200, replace = TRUE)
)

# Basic comparison analysis
comparison_results <- compare_groups(
  data = treatment_data,
  category = "group",
  Vars = c("anxiety", "depression", "stress"),
  plots = TRUE,
  table = TRUE
)

# View results
comparison_results$plots$group_vs_anxiety
comparison_results$summary_table

# Stratified analysis by study site
stratified_results <- compare_groups(
  data = treatment_data,
  category = "group",
  Vars = c("anxiety", "depression"),
  repeat_category = "study_site",
  plots = TRUE,
  table = TRUE
)

Advanced Features

Multiple Stratification Variables

Stratify by multiple variables simultaneously using c():

# Stratify by both study_site AND gender
multi_stratified <- compare_groups(
  data = treatment_data,
  category = "group",
  Vars = c("anxiety", "depression"),
  repeat_category = c("study_site", "gender"),
  table = TRUE
)

# Results are organized by combined groups (e.g., "Site_A | Male")

Filtering Specific Levels

Use repeat_levels to analyze only specific subgroups:

# Analyze only Site_A and Site_B
filtered_results <- compare_groups(
  data = treatment_data,
  category = "group",
  Vars = c("anxiety", "stress"),
  repeat_category = "study_site",
  repeat_levels = c("Site_A", "Site_B"),
  table = TRUE
)

Post-hoc Comparison Table

Generate a formatted table of all pairwise comparisons:

# Generate sample data with 3+ groups
set.seed(123)
multi_group_data <- data.frame(
  condition = sample(c("A", "B", "C"), 150, replace = TRUE),
  score = rnorm(150, mean = 50, sd = 10),
  performance = rnorm(150, mean = 75, sd = 15)
)

# Get post-hoc comparison table
posthoc_results <- compare_groups(
  data = multi_group_data,
  category = "condition",
  Vars = c("score", "performance"),
  posthoc = TRUE,
  posthoc_table = TRUE,
  table = TRUE
)

# Access the formatted post-hoc table
posthoc_results$posthoc_table

Nonparametric Tests

Use type shortcuts for different test types:

# Mann-Whitney U test (2 groups)
mw_results <- compare_groups(
  data = treatment_data,
  category = "group",
  Vars = c("anxiety", "depression"),
  type = "np",  # nonparametric
  table = TRUE
)

# Kruskal-Wallis test (3+ groups)
kw_results <- compare_groups(
  data = multi_group_data,
  category = "condition",
  Vars = c("score", "performance"),
  type = "np",
  table = TRUE
)

Key Features

Automatic test selection: t-test (2 groups) or ANOVA (3+ groups)
Post-hoc tests: Tukey HSD or Games-Howell for multiple comparisons
Effect sizes: Cohen’s d, eta-squared, epsilon-squared, rank-biserial
Multiple stratification: Use repeat_category = c("var1", "var2") for combined grouping
Level filtering: Use repeat_levels to select specific subgroups
Post-hoc table: Use posthoc_table = TRUE for formatted pairwise comparisons
Type shortcuts: type = "np" for nonparametric (Mann-Whitney/Kruskal-Wallis)

2. Correlation Matrix

The correlation_matrix() function creates publication-ready correlation tables with significance stars and optional heatmap visualization. Supports bivariate, partial, and semi-partial correlations.

# Basic correlation matrix (all numeric variables)
correlation_matrix(mtcars)

# Specific variables with title
correlation_matrix(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp", "hp", "wt"),
  title = "Motor Trend Car Correlations"
)

# With confidence intervals
correlation_matrix(
  data = mtcars,
  Vars = c("mpg", "hp", "wt"),
  show_ci = TRUE
)

# Partial correlations (controlling for other variables)
correlation_matrix(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp", "hp"),
  type = "partial",
  title = "Partial Correlations"
)

# Spearman correlations with Bonferroni correction
correlation_matrix(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp"),
  method = "spearman",
  p_adjust = "bonferroni"
)

# Full matrix with heatmap
result <- correlation_matrix(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp", "hp"),
  triangle = "full",
  heatmap = TRUE
)

# Access the heatmap plot
result$heatmap

3. Full Correlation Tables

The correlations() function creates comprehensive long-format correlation tables with full statistics for each variable pair. This is ideal for detailed reporting and when you need complete statistics (r, CI, t, df, p, n) for each correlation.

Basic Usage

# Full correlation table - all numeric variables
correlations(mtcars)

# Specify variables
correlations(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp", "hp", "wt")
)

# Only show strong correlations (|r| >= 0.7)
correlations(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp", "hp", "wt"),
  min_r = 0.7
)

# Only show significant correlations
correlations(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp", "hp", "wt", "qsec"),
  sig_only = TRUE
)

# With p-value adjustment
correlations(
  data = mtcars,
  Vars = c("mpg", "cyl", "disp", "hp"),
  p_adjust = "bonferroni"
)

Stratified Correlations by Group

Compute correlations separately for each group level, with results combined into one table:

# Correlations by number of cylinders
correlations(
  data = mtcars,
  Vars = c("mpg", "hp", "wt"),
  group_by = "cyl"
)

# Correlations by transmission type, only significant
correlations(
  data = mtcars,
  Vars = c("mpg", "hp", "wt", "disp"),
  group_by = "am",
  sig_only = TRUE
)

Multilevel Correlations

For repeated measures or hierarchically nested data, use multilevel correlations to estimate within-cluster associations while removing between-cluster variance:

# Create sample longitudinal data
set.seed(123)
n_subjects <- 30
n_timepoints <- 4

longitudinal_data <- data.frame(
  participant_id = rep(1:n_subjects, each = n_timepoints),
  time = rep(1:n_timepoints, n_subjects),
  anxiety = rnorm(n_subjects * n_timepoints, mean = 5, sd = 1.5),
  depression = rnorm(n_subjects * n_timepoints, mean = 4, sd = 1.2),
  stress = rnorm(n_subjects * n_timepoints, mean = 6, sd = 1.8)
)

# Within-person correlations (multilevel)
correlations(
  data = longitudinal_data,
  Vars = c("anxiety", "depression", "stress"),
  multilevel = TRUE,
  id = "participant_id"
)

# Include between-person correlations
correlations(
  data = longitudinal_data,
  Vars = c("anxiety", "depression"),
  multilevel = TRUE,
  id = "participant_id",
  between = TRUE
)

4. Data Transformation

The package provides consistent transformation functions with group-wise support.

Mean Centering

# Grand-mean centering
mtcars_c <- center(mtcars, Vars = c("mpg", "hp", "wt"))

# Group-mean centering
mtcars_c <- center(mtcars, Vars = c("mpg", "hp"), group_by = "cyl")

# With dplyr (vectorized version)
library(dplyr)
mtcars %>% mutate(mpg_c = center_vec(mpg))
mtcars %>% group_by(cyl) %>% mutate(mpg_c = center_vec(mpg))

Z-Score Standardization

# Grand standardization
mtcars_z <- standardize(mtcars, Vars = c("mpg", "hp", "wt"))

# Group-wise standardization
mtcars_z <- standardize(mtcars, Vars = c("mpg", "hp"), group_by = "cyl")

# With dplyr
mtcars %>% mutate(mpg_z = standardize_vec(mpg))

Scaling

# Scale by SD
mtcars_s <- scale_vars(mtcars, Vars = c("mpg", "hp"), method = "sd")

# Min-max scaling (0-1)
mtcars_s <- scale_vars(mtcars, Vars = c("mpg", "hp"), method = "range")

# Scale to custom range (1-10)
mtcars_s <- scale_vars(mtcars, Vars = c("mpg"), method = "range", range = c(1, 10))

Reverse Coding

For Likert scales and similar measures:

# Create sample data
df <- data.frame(
  item1 = c(1, 2, 3, 4, 5),
  item2 = c(5, 4, 3, 2, 1),  # reverse-worded
  item3 = c(2, 3, 4, 3, 2)
)

# Reverse code with auto-detected scale
df <- reverse_code(df, Vars = "item2")

# Explicit 1-5 Likert scale
df <- reverse_code(df, Vars = c("item2", "item3"), min = 1, max = 5)

# With dplyr
df %>% mutate(item2_r = reverse_code_vec(item2, min = 1, max = 5))

5. Descriptive Statistics

The descriptive_table() function creates publication-ready descriptive statistics tables.

# Generate sample data
data <- data.frame(
  age = rnorm(100, mean = 35, sd = 10),
  score = rnorm(100, mean = 75, sd = 15),
  income = rnorm(100, mean = 50000, sd = 15000),
  gender = sample(c("Male", "Female"), 100, replace = TRUE)
)

# Basic descriptive table
descriptive_table(
  data = data,
  Vars = c("age", "score", "income")
)

# With grouping and extended statistics
descriptive_table(
  data = data,
  Vars = c("age", "score"),
  group_by = "gender",
  stats = c("n", "mean", "sd", "median", "iqr", "skewness"),
  title = "Sample Characteristics",
  theme = "colorful"
)

6. Categorical Frequency Tables

The categorical_table() function creates frequency tables with chi-square tests.

# Generate sample data
data <- data.frame(
  gender = sample(c("Male", "Female", "Other"), 200, replace = TRUE),
  education = sample(c("High School", "Bachelor", "Master", "PhD"), 200, replace = TRUE)
)

# Single variable frequency table
categorical_table(data, var = "gender")

# Cross-tabulation with chi-square test
categorical_table(
  data = data,
  var = "gender",
  by = "education",
  chi_square = TRUE,
  cramers_v = TRUE
)

7. Missing Data Analysis

Comprehensive Analysis

# Add some missing values
df <- mtcars
df$mpg[c(1, 5, 10)] <- NA
df$hp[c(2, 5, 15)] <- NA

# Run missing data analysis
results <- missing_analysis(df)

# Components
results$summary      # gt table with per-variable missing counts
results$patterns     # Missing data patterns
results$mcar         # Little's MCAR test results
results$plot         # Missing pattern visualization

Imputation

# Replace with mean
df_imp <- replace_missing(df, Vars = "mpg", method = "mean")

# Replace with median
df_imp <- replace_missing(df, Vars = "mpg", method = "median")

# Group-wise imputation
df_imp <- replace_missing(df, Vars = "mpg", method = "mean", group_by = "cyl")

8. Latent Profile Clustering

clustering() fits latent profile models with the latents package, one for every combination of profile count and covariance structure, and keeps the one with the lowest BIC.

set.seed(123)
vars <- c("Sepal.Length", "Sepal.Width", "Petal.Length", "Petal.Width")
fit <- clustering(iris, vars, n_profiles = 2:4, models = c("EEE", "VVI"))
fit

summary() compares every candidate; as.data.frame() gives the profile means and fitted() the input data with each row’s profile.

summary(fit)
as.data.frame(fit)
head(fitted(fit))

plot() draws the profiles; type selects other views, listed by clustering_plot_types().

plot(fit)
plot(fit, type = c("heatmap", "avepp", "enumeration"))

9. Categorical Variable Analysis

The mosaic_analysis() function performs comprehensive analysis of categorical variables.

# Generate sample demographic data
set.seed(789)
demographic_data <- data.frame(
  gender = sample(c("Male", "Female", "Other"), 500, replace = TRUE, prob = c(0.45, 0.45, 0.1)),
  education = sample(c("High School", "Bachelor", "Master", "PhD"), 500, replace = TRUE),
  region = sample(c("North", "South", "East", "West"), 500, replace = TRUE)
)

# Perform mosaic analysis
mosaic_results <- mosaic_analysis(
  data = demographic_data,
  var1 = "gender",
  var2 = "education",
  min_count = 10,
  title = "Gender Distribution by Education Level",
  show_percentages = TRUE
)

# Access results
print(mosaic_results$chi_test)
print(mosaic_results$cramers_v)
print(mosaic_results$consolidated_table)

10. Network Estimation

The estimate_single_network() function estimates psychological networks.

# Generate sample data
set.seed(123)
network_data <- data.frame(
  var1 = rnorm(200),
  var2 = rnorm(200),
  var3 = rnorm(200),
  var4 = rnorm(200),
  var5 = rnorm(200)
)

# Estimate network
network_results <- estimate_single_network(
  data = network_data,
  vars = c("var1", "var2", "var3", "var4", "var5"),
  method = "EBICglasso"
)

# View results
print(network_results)
summary(network_results)

Best Practices

Data Preparation

Clean your data before analysis
Check for missing values and handle them appropriately
Verify variable types (numeric for clustering, categorical for mosaic)
Consider sample sizes - larger samples provide more reliable results

Parameter Selection

Clustering: Start with 2-5 clusters and use domain knowledge
Scaling: Use “standardize” for most cases
Minimum counts: Set appropriate thresholds based on your data size

Interpretation

Effect sizes: Consider effect sizes, not just p-values
Multiple testing: Use p-value adjustments when appropriate
Practical significance: Look beyond statistical significance

Getting Help

Use ?function_name for detailed documentation
Check the package vignettes for examples
Report issues on the GitHub repository

Conclusion

The Saqrmisc package provides a comprehensive toolkit for data analysis and visualization. The functions are designed to work together seamlessly while also being useful independently.

For more information, visit the package documentation or GitHub repository.