Saqrmisc

A comprehensive R package for statistical analysis, data transformation, and visualization with publication-ready outputs. All functions use a consistent quoted-string API and support multiple output formats (gt, plain, markdown, latex, kable).

Table of Contents

Installation

# devtools::install_github("mohsaqr/Saqrmisc")
install.packages("Saqrmisc", repos = c('https://mohsaqr.r-universe.dev', 'https://cloud.r-project.org'))
library(Saqrmisc)

Dependencies:

install.packages(c(
  "dplyr", "ggplot2", "gt", "tibble", "tidyr", "rlang",
  "MoEClust", "mclust", "vcd", "gridExtra",
  "bootnet", "qgraph", "mgm",
  "BayesFactor", "effectsize", "effsize"
))

# Optional
install.packages(c("ggstatsplot", "TOSTER", "kableExtra", "httr2"))

Correlation Analysis

correlations()

Full pairwise correlation table with coefficients, CIs, t-statistics, df, p-values, and sample sizes.

correlations(data, Vars = NULL, type = "bivariate", method = "pearson",
             p_adjust = "none", ci_level = 0.95, min_r = NULL, sig_only = FALSE,
             multilevel = FALSE, id = NULL, between = FALSE, group_by = NULL,
             include = NULL, exclude = NULL, digits = 3, format = "gt")
Parameter Description Default
Vars Variables to correlate (NULL = all numeric) NULL
type "bivariate", "partial", "semi-partial" "bivariate"
method "pearson", "spearman", "kendall" "pearson"
p_adjust "none", "bonferroni", "holm", "fdr" "none"
min_r Show only correlations above this threshold NULL
sig_only Show only significant correlations FALSE
multilevel Within-cluster correlations (requires id) FALSE
group_by Stratified correlations by group NULL 
correlations(mtcars, Vars = c("mpg", "hp", "wt"))
correlations(mtcars, Vars = c("mpg", "hp", "wt"), type = "partial", sig_only = TRUE)

Returns: List with $table, $data, $n_pairs, $n_significant.

correlation_matrix()

Publication-ready correlation matrix with significance stars and optional heatmap.

correlation_matrix(data, Vars = NULL, type = "bivariate", method = "pearson",
                   triangle = "lower", diagonal = "dash", show_n = FALSE,
                   show_ci = FALSE, show_p = FALSE, p_adjust = "none",
                   stars = TRUE, heatmap = FALSE, digits = 2, format = "gt")
Parameter Description Default
triangle "lower", "upper", "full" "lower"
show_ci Display confidence intervals FALSE
show_n Display pairwise sample sizes FALSE
heatmap Generate heatmap visualization FALSE 
correlation_matrix(mtcars, Vars = c("mpg", "hp", "wt", "qsec"), show_ci = TRUE)

Returns: List with $table, $correlation_matrix, $p_matrix, $heatmap.

Group Comparisons

compare_groups()

Compare groups using t-tests, ANOVA, or nonparametric alternatives with publication-ready tables and plots. Automatically selects the appropriate test based on group count and data properties.

compare_groups(data, category, Vars = NULL, repeat_category = NULL,
               compare_by = NULL, compare_mode = "between", type = "auto",
               posthoc = TRUE, posthoc_method = "games-howell",
               bayesian = FALSE, equivalence = FALSE, equivalence_bounds = NULL,
               pivot = FALSE, pivot_stat = "mean", format = "gt")
Parameter Description Default
category Grouping variable (required) --
Vars Numeric variables to compare (NULL = all) NULL
type "auto", "p" (parametric), "np" (nonparametric), "bayes" "auto"
posthoc Compute post-hoc tests (3+ groups) TRUE
posthoc_method "games-howell" or "tukey" "games-howell"
bayesian Compute Bayes Factors FALSE
equivalence Perform TOST equivalence testing FALSE
repeat_category Stratify analysis by this variable NULL
pivot Compact table with categories as columns FALSE
pivot_stat "mean", "mean_sd", "median", "n" "mean"
compare_by Factors to test within each category NULL
compare_mode "between" or "within" "between" 
# Two-group comparison (t-test)
results <- compare_groups(mtcars, category = "am", Vars = c("mpg", "hp"))

# ANOVA with post-hoc
results <- compare_groups(mtcars, category = "cyl", Vars = c("mpg", "hp"))

Returns: List with $summary_table, $summary_data, $plots, $pivot_table, $within_table.

Pivot Tables

Compact summary with category levels as columns and a p (ES) column:

results <- compare_groups(df, category = "group", Vars = c("x", "y"),
                          pivot = TRUE, pivot_stat = "mean_sd")
results$pivot_table

Within-Group Analysis

Test whether factors have effects within each category level:

results <- compare_groups(df, category = "llm", Vars = "score",
                          compare_by = c("pronoun", "support"),
                          compare_mode = "within")
results$within_table

Descriptive Statistics

descriptive_table()

Flexible summary statistics table with selectable statistics.

descriptive_table(data, Vars = NULL,
                  stats = c("n", "mean", "sd", "median", "min", "max"),
                  group_by = NULL, overall = FALSE, labels = NULL,
                  digits = 3, format = "gt")
Parameter Description Default
stats Statistics to include (see below) c("n","mean","sd","median","min","max")
group_by Compute statistics by group NULL
overall Include overall/total row FALSE
labels Named vector of variable labels NULL

Available statistics: n, missing, missing_pct, mean, sd, se, var, median, min, max, range, iqr, q1, q3, skewness, kurtosis, cv

descriptive_table(mtcars, Vars = c("mpg", "hp", "wt"), group_by = "cyl", overall = TRUE)

categorical_table()

Frequency and cross-tabulation tables with optional chi-square test.

categorical_table(data, var, by = NULL, chi_square = FALSE,
                  cramers_v = FALSE, group_by = NULL, format = "gt")

categorical_table(mtcars, var = "cyl", by = "am", chi_square = TRUE, cramers_v = TRUE)

auto_describe()

Automatically detects variable types and produces appropriate summaries for each.

auto_describe(data, group_by = NULL, format = "gt")

Returns: List with $numeric, $categorical, $variable_types.

Data Transformation

All transformation functions add new columns with a configurable suffix and support group-wise operations via group_by. Each has a _vec() variant for use in dplyr::mutate().

center()

Mean centering (grand-mean or group-mean).

center(data, Vars, suffix = "_c", group_by = NULL)

center(mtcars, Vars = c("mpg", "hp"), group_by = "cyl")

standardize()

Z-score standardization.

standardize(data, Vars, suffix = "_z", group_by = NULL)

scale_vars()

Scale by standard deviation or rescale to a custom range.

scale_vars(data, Vars, method = "sd", range = c(0, 1), suffix = "_s", group_by = NULL)
Parameter Description Default
method "sd" (divide by SD) or "range" (min-max) "sd"
range Target range for "range" method c(0, 1)

reverse_code()

Reverse code items for Likert scales. Auto-detects or accepts explicit min/max.

reverse_code(data, Vars, min = NULL, max = NULL, suffix = "_r")

reverse_code(df, Vars = c("item2", "item5"), min = 1, max = 5)

Missing Data

missing_analysis()

Comprehensive missing data diagnostics including patterns, Little's MCAR test, and visualizations.

missing_analysis(data, Vars = NULL, pattern_plot = TRUE,
                 mcar_test = TRUE, correlations = FALSE, digits = 2)

Returns: List with $summary, $patterns, $mcar, $plot.

replace_missing()

Simple imputation by mean or median, with optional group-wise computation.

replace_missing(data, Vars, method = "mean", group_by = NULL, suffix = "_imp")
Parameter Description Default
method "mean" or "median" "mean"
suffix Suffix for new column (NULL to replace in-place) "_imp"

Outlier Analysis

outlier_check()

Detect outliers using multiple methods.

outlier_check(data, Vars, method = "zscore", threshold = 3,
              flag = TRUE, plot = TRUE)
Parameter Description Default
method "zscore", "iqr", "percentile", "mahalanobis" "zscore"
threshold Detection threshold (method-dependent) 3

Returns: List with $summary, $data, $outlier_indices, $plot.

replace_outliers()

Treat outliers by winsorizing, replacing with NA, or substituting with median/mean.

replace_outliers(data, Vars, method = "winsorize", threshold = 0.05,
                 suffix = "_treated")
Parameter Description Default
method "winsorize", "na", "median", "mean" "winsorize"

Normality Testing

normality_check()

Shapiro-Wilk test, skewness, kurtosis, Q-Q plots, and histograms.

normality_check(data, Vars, digits = 3)

Returns: List with $summary, $plots, $overall.

Clustering

clustering()

Latent profile clustering with the latents package. Fits every combination of profile counts and covariance structures (the 14 mclust codes, EII through VVV) and keeps the one with the lowest BIC.

clustering(data, vars, n_profiles, models = "VVI",
           scaling = "standardize", n_starts = 10, seed = NULL,
           na_action = "omit", verbose = TRUE)
Parameter Description Default
n_profiles Profile count(s), e.g. 3 or 2:5 (alias n_clusters) required
models Covariance codes, e.g. c("EEE", "VVI"), or "all" "VVI"
scaling "standardize", "center", "minmax", "none" "standardize" 
fit <- clustering(mtcars, vars = c("mpg", "hp", "wt"), n_profiles = 2:4,
                  models = c("EEE", "VVI"))
fit                    # the selected model
summary(fit)           # every candidate, with delta BIC
as.data.frame(fit)     # profile means
fitted(fit)            # input data with profile and posteriors
plot(fit)              # profile plots; see clustering_plot_types()
plot(fit, type = "diagnostics")

The result is a latents fit, so latents::diagnostics(fit), latents::get_results(fit, what) and predict() work on it directly.

Network Analysis

estimate_single_network()

Estimate psychological networks using EBICglasso regularized partial correlations.

estimate_single_network(df, Vars = NULL, layout = "circle",
                        default = "EBICglasso", verbose = TRUE,
                        compute_centrality = TRUE)

Returns: Object with $plot, $centrality, $edges, $network_obj.

compare_networks()

Estimate and compare networks between groups.

compare_networks(data, group_var, variables = NULL,
                 method = "EBICglasso", layout = "circle")

Categorical Analysis

mosaic_analysis()

Mosaic plot with chi-square test and Cramer's V.

mosaic_analysis(data, var1, var2, show_percentages = TRUE,
                percentage_base = "total")

Returns: Object with $plot, $chi_test, $cramers_v, $summary.

AI Interpretation

pass()

Pipe any R output to an AI model for scientific interpretation.

pass(result, action = "interpret", style = "scientific", output = "text",
     provider = "openai", model = NULL, context = NULL,
     system_message = NULL, prompt = NULL, base_url = NULL)
Parameter Options Default
action "interpret", "explain", "write", "summarize", "critique", "suggest" "interpret"
style "scientific", "simple", "detailed", "brief" "scientific"
output "text", "markdown", "latex", "html" "text"
provider "openai", "anthropic", "gemini", "openrouter", or use base_url for local "openai" 
t.test(mpg ~ am, data = mtcars) |> pass()
lm(mpg ~ wt + hp, data = mtcars) |> summary() |> pass(action = "write")

API Key Setup

set_openai_key("sk-...")
set_claude_key("sk-ant-...")
set_gemini_key("AIza...")
set_openrouter_key("sk-or-...")

# Or via environment variables in .Renviron:
# OPENAI_API_KEY=your-key
# ANTHROPIC_API_KEY=your-key

Output Formats

All table-producing functions share these parameters:

Parameter Options Default
format "gt", "plain", "markdown", "latex", "kable" "gt"
show_header TRUE / FALSE TRUE

Results can also be converted after the fact with to_gt(), to_markdown(), to_latex(), to_html(), to_kable(), and to_dataframe().

Vectorized Functions

For use in dplyr::mutate() and across() workflows:

Function Description
center_vec(x) Mean-center a vector
standardize_vec(x) Z-score standardize
scale_vec(x, method, range) Scale by SD or range
reverse_code_vec(x, min, max) Reverse code
winsorize_vec(x, probs) Cap extreme values
is_outlier(x, method, threshold) Logical outlier flag

These respect dplyr::group_by() for group-wise operations.

Citation

Saqr, M. (2025). Saqrmisc: Comprehensive Data Analysis and Visualization Tools for R.
GitHub: https://github.com/mohsaqr/Saqrmisc

License

MIT License - see LICENSE for details.

Author

Mohammed Saqr - saqr@saqr.me - @mohsaqr