agriDQ 0.1.0

First CRAN release

New functions

• check_outliers() — univariate outlier detection (IQR fence, Z-score, Hampel identifier, Grubbs, Dixon Q-test) with consensus flagging. S3 print and plot methods included.
• check_outliers_mv() — Mahalanobis distance multivariate outlier detection with optional robust (MCD) covariance and ridge regularisation for near-singular matrices.
• check_missing() — missing data analysis with Little’s MCAR test (internal implementation), pattern matrix, mechanism classification, and missingness heatmap.
• classify_missing() — per-variable logistic regression to classify missingness as MCAR, MAR, or MNAR.
• check_normality() — battery of six normality tests (Shapiro-Wilk, Anderson-Darling, Kolmogorov-Smirnov, Lilliefors, Pearson chi-square, Jarque-Bera) with skewness/kurtosis diagnostics and consensus recommendation. Jarque-Bera is implemented internally with no external dependency.
• check_homogeneity() — Bartlett, Levene (Brown-Forsythe, center = median), and Fligner-Killeen tests for homogeneity of variance with practical variance ratio.
• check_independence() — Durbin-Watson, Breusch-Godfrey, and Wald-Wolfowitz runs tests for independence of residuals, with ACF plot.
• check_design() — experimental design validation for CRD, RCBD, LSD, and factorial designs: balance, completeness, error df (Gomez & Gomez guideline), and missing treatment combinations.
• check_qualitative() — categorical variable quality checks: case inconsistency, whitespace, near-duplicate labels (Levenshtein, long labels only), unexpected levels, and rare categories.
• standardise_labels() — automatic label standardisation with case conversion and lookup-table replacement.
• run_dq_pipeline() — single-call full pipeline runner returning a master summary data frame.
• generate_dq_report() — automated self-contained HTML report with green/amber/red scorecard.

Data

• agri_trial — simulated RCBD wheat variety trial (20 plots, 4 treatments T1-T4, 5 blocks B1-B5) with one intentional outlier and one missing value for demonstration.

Notes

• All skewness, kurtosis, and Jarque-Bera calculations are implemented internally, eliminating the dependency on the moments package.
• Grubbs and Dixon Q-tests are implemented internally, eliminating the dependency on the outliers package.
• Little’s MCAR test is implemented internally, eliminating the dependency on BaylorEdPsych or MissMech.
• The package requires R >= 4.1.0.