pyrsm

Create custom plots from a polars DataFrame in Python using the pyrsm library's `visualize` function — a plotnine wrapper supporting 8 geom types (dist / hist / density / scatter / bar / line / box / violin) with full control over aesthetics (color / fill / shape / group / linetype), faceting (facet_wrap / facet_grid), aggregation, smoothing (lm / loess), jitter, titles, and multiple x/y variables. The return value is a `plotnine.ggplot` object for one plot or a plotnine composition for multiple plots, and users can extend the result with plotnine layers — scales, themes, geom_smooth, geom_vline, custom labels — making plotnine extensibility a first-class part of this skill. Triggers include phrases like "scatter plot of x vs y", "histogram of price", "box plot of price by cut", "line chart over time, colored by region", "facet by category", "add a regression line", "use plotnine to plot", "density plot of carat by cut", or any request for a specific plot in a marketing/business analytics context.

Combine two polars DataFrames in Python using the pyrsm library's `combine` function — supporting inner / left / right / full / semi / anti joins, vertical and horizontal binds, and set operations (intersect / union / setdiff). Use this skill whenever a student or analyst wants to merge two tables on a key, stack two tables of the same schema, identify rows that appear in both tables (or in one but not the other), or work with the pyrsm package for any two-DataFrame combining task. The output is a polars DataFrame the user can keep chaining onto. Triggers include phrases like "join these two tables", "merge on customer_id", "left join", "stack two datasets", "concatenate rows", "find rows in A but not in B", "anti join to find unmatched", "set intersection of two tables", "bind columns side by side", or any mention of combining two tabular datasets in a marketing/business analytics context.

Compare means of a numeric variable across two or more groups in Python using the pyrsm library's `compare_means` class. Use this skill whenever a student or analyst wants to run a two-sample t-test, a paired t-test, or an ANOVA-style set of pairwise comparisons; switch to a Wilcoxon rank-sum / signed-rank test for skewed or small samples; apply Bonferroni adjustment to control family-wise error across multiple comparisons; choose between independent and paired samples; or work with the pyrsm package for any group-mean-comparison task — even if they don't explicitly say "pyrsm" or "compare_means". Triggers include phrases like "compare means across groups", "t-test between two groups", "ANOVA across professor ranks", "is salary different by sex/discipline", "before-after paired test", "Wilcoxon rank-sum test", "Bonferroni adjustment", "pairwise comparisons with multiple-testing correction", or any mention of comparing a continuous outcome across categorical groups in a marketing/business analytics class.

Compare proportions of a categorical outcome across two or more groups in Python using the pyrsm library's `compare_props` class. Use this skill whenever a student or analyst wants to test if the proportion of a "success" level (yes / churned / clicked / survived) differs across groups defined by another categorical variable, focus on specific pair comparisons via `comb`, apply Bonferroni adjustment, validate the chi-squared cell-count assumption, or work with the pyrsm package for any group-proportion-comparison task — even if they don't explicitly say "pyrsm" or "compare_props". Triggers include phrases like "compare proportions across groups", "does survival rate differ by class", "did the click-rate change between A and B and C", "two-proportion z-test", "is the yes-rate the same for males and females", "pairwise proportion comparisons with adjustment", or any mention of comparing a binary outcome across two or more categorical groups in a marketing/business analytics class.

Compute and interpret correlation matrices in Python using the pyrsm library's `correlation` class. Use this skill whenever a student or analyst wants to measure the strength and direction of pairwise relationships among numeric variables, choose between Pearson, Spearman, Kendall, or polychoric correlation, build a scatter-matrix visualization with significance stars, screen for multicollinearity in a dataset before fitting a regression, or work with the pyrsm package for any correlation task — even if they don't explicitly say "pyrsm" or "correlation". Triggers include phrases like "how correlated are X and Y", "compute a correlation matrix", "is there a relationship between these variables", "check for multicollinearity", "Pearson vs Spearman", "what's the rank correlation", "scatter matrix with regression lines", or any mention of measuring pairwise linear/monotonic association among continuous variables in a marketing/business analytics class.

Build and interpret two-way contingency tables and chi-squared tests of independence in Python using the pyrsm library's `cross_tabs` class. Use this skill whenever a student or analyst wants to test whether two categorical variables are associated, build observed/expected/chi-squared/standardized-deviation tables, identify which cells drive an omnibus rejection, compute row/column/total percentages, validate the expected-cell-count assumption, or work with the pyrsm package for any contingency-table task — even if they don't explicitly say "pyrsm" or "cross_tabs". Triggers include phrases like "is X associated with Y", "chi-square test of independence", "cross-tab", "contingency table", "does income predict newspaper choice", "which cells contribute to the chi-squared", "standardized residuals", "expected vs observed frequencies", "row percentages vs column percentages", or any mention of analyzing the relationship between two categorical variables in a marketing/business analytics class.

Run distribution analysis on a polars DataFrame in Python using the pyrsm library's `distr` class. Use this skill whenever a student or analyst wants a one-shot exploratory summary of every column in a dataset — numeric variables get mean/median/min/max/sd, categorical variables get counts and proportions per level, other (date/datetime) get min/max/n_unique/n_missing — paired with histograms (numeric) and bar plots (categorical). The class also auto-classifies low-cardinality integers as categorical, and returns plotnine ggplot objects users can layer additional aesthetics onto. Triggers include phrases like "describe this dataset", "show me the distributions", "plot all the variables", "categorical vs numeric split", "histograms for every numeric column", "what's in this DataFrame", "EDA dump", or any request for a holistic first-look at a dataset in a marketing/business analytics context.

Build, validate, solve, and interpret decision trees with pyrsm.model.dtree

Compute summary statistics for numeric columns in Python using the pyrsm library's `explore` function. Use this skill whenever a student or analyst wants a quick numeric summary (mean, median, min, max, sd, count, missing, n_unique) across one or many columns of a polars DataFrame, optionally grouped by a categorical column, with categorical predictors auto-expanded to dummies. The output is itself a polars DataFrame that the user can further chain with `.filter`, `.sort`, `.with_columns` — extensibility is a first-class part of this skill. Triggers include phrases like "describe these variables", "summary statistics", "group means by category", "what's the mean and median of price by cut", "give me a quick describe", "explore this dataset", "show me missingness per column", "n_unique per variable", "count rows per group with mean and sd", or any mention of wanting per-column numeric summaries in a marketing/business analytics context.

Run and interpret chi-squared goodness-of-fit tests in Python using the pyrsm library's `goodness` class. Use this skill whenever a student or analyst wants to compare the observed distribution of a single categorical variable to an expected distribution (uniform, census-based, historical, or theoretical), check which cells drive any deviation via standardized residuals, or work with the pyrsm package for any one-variable chi-squared task — even if they don't explicitly say "pyrsm" or "goodness". Triggers include phrases like "test if this categorical variable follows expected proportions", "chi-square goodness-of-fit", "does the income split match the census 70/30", "is the sample evenly distributed across categories", "are the levels of X consistent with a hypothesized distribution", "test if dice are fair", "compare observed vs expected frequencies", or any mention of testing a single discrete distribution against a benchmark in a marketing/business analytics class.

Run and interpret logistic regression analyses in Python using the pyrsm library's `logistic` class. Use this skill whenever a student or analyst wants to fit a logistic or binary classification model, predict the probability of a binary or categorical outcome, interpret odds ratios, evaluate AUC or pseudo R-squared, compare predictor importance, generate prediction scenarios, or work with the pyrsm package for any classification task — even if they don't explicitly say "pyrsm" or "logistic". Triggers include phrases like "predict whether a customer will churn", "model a binary outcome", "what predicts survival", "fit a classification model", "interpret odds ratios", "which variables predict purchase", "binary dependent variable", "are these odds ratios significant", or any mention of modeling a yes/no, 0/1, or two-level categorical response in a business or analytics context.

Fit and interpret a Multi-Layer Perceptron (feed-forward neural network) in Python using the pyrsm library's `mlp` class — for either binary classification (`mod_type="classification"`, predict P(`lev`)) or regression (`mod_type="regression"`, predict a continuous outcome). Use this skill whenever a student or analyst wants to fit a neural network on tabular data, tune hidden layer sizes / regularization (alpha) / solver with cross-validation, examine feature importance (permutation), look at partial-dependence plots, score new data, or evaluate classification performance (confusion, AUC, gains, lift, profit). The class automatically z-scores features before training (NN models require scaling). Triggers include phrases like "fit a neural network", "MLP classifier", "tune hidden layers", "neural network feature importance", "compare neural net vs logistic", "PDP for MLP", "neural net for tabular data", or any feed-forward NN modeling request in a marketing/business analytics context.

Build pivot tables, frequency tables, and crosstabs in Python using the pyrsm library's `pivot` function. Use this skill whenever a student or analyst wants to count occurrences of categorical levels, cross-tabulate two categorical variables, compute mean/median/sum/sd of a numeric value across grouping variables, normalize a crosstab into row/column/total percentages, add row and column totals, or work with polars-first wide-format outputs. The output is a polars DataFrame, so it can be further chained with any polars expression — extensibility is a first-class part of this skill. Triggers include phrases like "make a frequency table", "crosstab cut by color", "pivot table of mean price by cut and color", "row percentages of survival by class", "count rows per group", "summarize sales by region and product", "category counts with totals", or any descriptive cross-tabulation request in a marketing/business analytics context.

Compute tail probabilities, critical values, expected values, and standard deviations for common probability distributions in Python using the pyrsm library's `prob_calc` class. Use this skill whenever a student or analyst wants to find a probability given a value, find a value given a probability, look up a critical t / z / chi-squared / F for a hypothesis test, compute service-level inventory thresholds, derive p-values from a test statistic, calculate the mean and variance of a discrete distribution, or work with the pyrsm package for any distributional calculation. Triggers include phrases like "what's the probability that …", "find the critical t at α=0.05", "what value of X gives 95% probability below it", "compute the p-value for chi-sq = 32.84 on 1 df", "binomial probability of 15 or more successes", "normal critical value for service level 95%", "Poisson tail probability", "uniform between a and b", "inventory level for 95% service", or any mention of looking up probabilities or values from the Binom

Run and interpret linear regression analyses in Python using the pyrsm library's `regress` class. Use this skill whenever a student or analyst wants to fit a linear regression, interpret coefficients, check residual diagnostics, run an F-test, compare linear vs log-log specifications, generate predictions, or work with the pyrsm package for an OLS task — even if they don't explicitly say "pyrsm" or "regress". Triggers include phrases like "regress y on x", "fit a linear model", "run an OLS", "interpret these coefficients", "are these residuals OK", "should I log-transform", or any mention of fitting a model to a parquet/csv dataset for a marketing/business analytics class.

Fit and interpret Random Forest models in Python using the pyrsm library's `rforest` class — for either binary classification (`mod_type="classification"`, predict P(`lev`)) or regression (`mod_type="regression"`, predict a continuous outcome). Use this skill whenever a student or analyst wants to fit a Random Forest, get an OOB AUC / R² baseline, examine feature importance (permutation or sklearn-style mean-decrease-impurity), look at partial-dependence plots, tune `n_estimators` / `max_features` / `min_samples_leaf` with cross-validation, score new data, or evaluate classification performance (confusion, AUC, gains, lift, profit). Triggers include phrases like "fit a random forest", "tune a random forest", "feature importance from random forest", "OOB AUC", "predict churn with rforest", "compare logistic vs random forest", "PDP for the random forest", "GridSearchCV with pyrsm.model.rforest", or any tree-ensemble modeling request in a marketing/business analytics context.

Build, run, and interpret Monte Carlo simulations with pyrsm.model.simulate

Run and interpret one-sample mean tests in Python using the pyrsm library's `single_mean` class. Use this skill whenever a student or analyst wants to compare a sample mean to a hypothesized population value, decide whether the difference is statistically meaningful, choose a one-sided vs two-sided alternative, build or read a confidence interval around a mean, or work with the pyrsm package for any one-sample-t-test task — even if they don't explicitly say "pyrsm" or "single_mean". Triggers include phrases like "is the average really X?", "test whether the mean equals/exceeds/is below some target", "one-sample t-test", "compare a sample mean to a benchmark", "does this column's mean differ from 1750", "build a 95% CI for the mean", or any mention of testing a single continuous variable against a target value in a marketing/business analytics class.

Run and interpret one-sample proportion tests in Python using the pyrsm library's `single_prop` class. Use this skill whenever a student or analyst wants to compare a sample proportion (a yes/no, success/failure, or two-level categorical outcome) to a hypothesized population proportion, choose between an exact binomial test and a z-test approximation, build or read a confidence interval for a proportion, decide whether brand preference / churn rate / response rate / defect rate differs from a benchmark, or work with the pyrsm package for any one-sample-proportion task — even if they don't explicitly say "pyrsm" or "single_prop". Triggers include phrases like "is the brand preference really 10%", "test whether response rate exceeds X", "one-sample binomial test", "build a 95% CI for a proportion", "compare a yes/no outcome to a benchmark", "what fraction of customers said yes vs the target", or any mention of testing a binary outcome's prevalence against a target value in a marketing/business analytics class.

Convert wide-format data into long format (melt / gather) in Python using the pyrsm library's `unpivot` function. Use this skill whenever a student or analyst has columns that should really be rows of a categorical (Q1/Q2/Q3/Q4 → quarter; 2019/2020/2021 → year; product_a/product_b → product), needs to reshape data for plotting with plotnine, wants to feed a wide pivot table back into a long-format analysis, or generally needs to make the data "tidy" for a downstream tool. The output is a polars DataFrame the user can keep chaining onto. Triggers include phrases like "wide to long", "melt this table", "unpivot the quarters", "convert columns Q1-Q4 to a single quarter column", "reshape for plotting", "make this tidy", "tidy data", or any mention of needing long-format data after producing a wide-format summary.

Fit and interpret XGBoost (gradient-boosted trees) models in Python using the pyrsm library's `xgboost` class — for either binary classification (`mod_type="classification"`, predict P(`lev`)) or regression (`mod_type="regression"`, predict a continuous outcome). Use this skill whenever a student or analyst wants to fit a gradient-boosted model, tune the learning_rate / n_estimators / max_depth / regularization tradeoff with cross-validation, examine feature importance (permutation or xgboost's built-in), look at partial-dependence plots, score new data, or evaluate classification performance (confusion, AUC, gains, lift, profit). Triggers include phrases like "fit an xgboost", "boosting model", "tune xgboost", "gradient boosted trees", "compare random forest vs xgboost", "what's the best learning rate", "xgboost feature importance", "PDP for xgboost", "GridSearchCV with pyrsm.model.xgboost", or any boosted-tree modeling request in a marketing/business analytics context.