The R script

Every analysis report that has R shows a Reproducibility (R) block with the emitted script, below the result. SPSS syntax is separate, under SPSS syntax (reference).

The script loads your download as:

df <- read.csv("dataset.csv", stringsAsFactors = FALSE, check.names = FALSE,
               na.strings = c('', 'NA', 'null', 'NULL'))

Identity comments at the top are # dataset_id: and # source: dataset.csv.

BadgesCopy Anchor Link

The R syntax badge is on engine analyses, not plugins. The live allow-list is _BADGE in app/r_coverage.py: when Rscript can run, those keys stamp verified or not_verified; without Rscript they can stamp verified_in_ci on this build.

not_verified is the amber state. The live statement on the badge can be more specific (for example ANOVA includes ΔF and Δn). Keys that are not in _BADGE stay unknown.

_BADGE keys: anova_oneway, anova_twoway, linear_regression, logistic_regression, descriptives, correlation, ttest_one_sample, ttest_paired, stepwise_regression, anova_mixed, mixed_anova, mann_whitney_u, wilcoxon_signed_rank, kruskal_wallis, friedman, sign_test, median_test, anova_repeated, rm_anova, anova_threeway, ancova, manova, hotelling_t2, hierarchical_regression, moderation_analysis, poisson_regression, probit_regression, negative_binomial_regression, ordinal_regression, partial_correlation, multilevel_modelling, fishers_exact, odds_ratio, relative_risk, mcnemar, cochrans_q, mantel_haenszel, cochran_armitage, cohens_kappa, weighted_kappa, fleiss_kappa, kendalls_w, goodman_kruskal_gamma, goodman_kruskal_lambda, somers_d, loglinear, jonckheere_terpstra, moses_test, runs_test, kolmogorov_smirnov, lilliefors_ks, shapiro_wilk, linear_mixed_model, mixed_model, ttest_independent, chi_square, reliability, reliability_cronbach, pca, kaplan_meier, cox_proportional_hazards, nelson_aalen, autocorrelation, nps, gee, generalized_linear_mixed_model, banner_table, batch_tables, rake, poststratify, merge_datasets, fuse_waves, set_active_weight, turf, maxdiff_count, van_westendorp, gabor_granger, funnel, drivers, choice_simulator, confirmatory_factor_analysis, structural_equation_modelling, efa, cluster_analysis, discriminant_analysis, canonical_correlation, correspondence, multidimensional_scaling, arima_sarima, exponential_smoothing, stl_decomposition, stationarity_tests, network, dbscan, latent_class_analysis, maxdiff_mnl, conjoint_mnl.

Not compared to R (NOT_VERIFIABLE in the same file)Copy Anchor Link

The badge stays off. Reasons from r_coverage.py:

• Open-text coding — an agent procedure, not a statistic with an R equivalent
• Decision tree — sklearn DecisionTree vs rpart splits differ; CART implementations are not aligned
• Random forest classification and regression — sklearn RandomForest vs randomForest; trees are not aligned across libraries
• Gradient boosting — sklearn GradientBoosting vs gbm; boosting paths differ
• SVM classification — sklearn SVC vs e1071::svm kernels/solvers differ
• Neural network (MLP) — sklearn MLP vs nnet; random init and architecture differ
• MaxDiff HB — penalized mixed-logit HB is not Sawtooth and has no unique R reference
• Conjoint HB — penalized mixed-logit CBC is not Sawtooth and has no unique R reference

How to run itCopy Anchor Link

1. Export the dataset as CSV (same names as the .sav, including weight when present).
2. Save the emitted script next to dataset.csv.
3. Run Rscript your-script.R. Tensr uses RSCRIPT_BIN if set, otherwise Rscript.

Packages the runner expects include jsonlite, dplyr, irr, epitools, DescTools, ez, ggplot2, and analysis-specific libraries such as car, lme4, MASS, survival, and geepack.

Open-text coding emits a comment only: it is not a stats engine call.