Best Jasper prompts for Statisticians

I have a dataset of customer purchase behavior with 20 features, including demographic information and purchase history. Using Python and scikit-learn, develop a step-by-step approach to select the most relevant features, handle missing values, and train a linear regression model to predict future purchases. Include a grid search for hyperparameter tuning and provide a statistical summary of the model's performance on a holdout set. Consider using techniques like cross-validation and regularization to prevent overfitting.

Given a complex SQL query that joins multiple large tables in Snowflake, analyze the query plan and identify bottlenecks in performance. Develop a rewritten query that uses efficient join techniques, such as using common table expressions or optimizing subqueries, and provide a comparison of the execution times before and after optimization. Consider using tools like Snowflake's query profiler to inform your optimization strategy.

Design an ETL pipeline using Python and pandas to extract data from a raw dataset containing customer feedback, clean and preprocess the data by handling missing values and outliers, and load the transformed data into a structured format for analysis. Include data quality checks and provide a statistical summary of the cleaned dataset, including summary statistics and data visualizations. Consider using techniques like data normalization and feature scaling to improve the quality of the data.

Compare the performance of two different machine learning models, a random forest and a gradient boosting model, on a classification task using a dataset with 10 features and 1000 samples. Using Python and scikit-learn, develop a step-by-step approach to train and tune each model, evaluate their performance using metrics like accuracy, precision, and recall, and provide a statistical summary of the results, including a comparison of the models' strengths and weaknesses. Consider using techniques like cross-validation and bootstrapping to estimate the models' performance on unseen data.