Machine Learning and Data Science in STATA

Learn how to import an Excel spreadsheet into Stata, set first row as variable names, inspect string and numeric variables, save the dataset, and capture commands in a do file.

Import data from an Excel spreadsheet into stata, use the first row as variable names, view a two-variable numeric dataset, and save via the command line for reproducible do-file workflow.

Explore the Stata interface, including the command line, menus, and data workflows from open to import and export, describe and manage variables, and generate graphics for machine learning.

Merge two data sets in Stata using the client ID key, handle 1-to-1, many-to-one, or many-to-many joins, and verify with summary statistics.

Combine multiple histograms in Stata using graph combine to compare distributions of loan disbursed, monthly income, client banking years, and work experience, and save graphs for reports.

Learn to create histograms in Stata, visualize distributions for continuous and discrete data, customize bins and axes, and save the commands in a do file for reuse.

Learn to edit histograms in Stata using the graph editor, tune graph properties, colors, margins, and titles, and save outputs as png for publications.

Visualize category data in Stata using bar charts and pie charts, comparing frequencies and percentage shares for gender (female vs male) and saving work with a do file.

Explore visualizing category data in stata by creating and combining bar charts for rural versus urban areas and educational grades, using frequency and percentage views to reveal distribution.

Explore how scatter plots reveal relationships between variables, including linear, parabolic, negative, and exponential patterns, and distinguish strong versus weak correlations. Learn to plot these relationships in Stata.

Create and compare scatter plots in Stata to explore monthly income versus work experience, filtering by gender, using two-way graphs, colors and symbols, and adding a fitted regression line.

Inspect data types and distributions in Stata using describe and inspect to quickly assess continuous variables, zeros, and patterns like exponential distribution.

Learn to use the sum command in STATA to summarize variables, report mean, standard deviation, min, max, percentiles, and skewness, and distinguish numeric from categorical data.

Use the encode command in Stata to convert string variables like gender and education into numeric codes, enabling descriptive statistics and preparation for regression and machine learning.

Compute weight of evidence for gender and other categories in STATA, using column percentages and the natural logarithm to gauge association, and prepare for logistic regression in this course.

Compute the weight of evidence for relationship status and education to select regression predictors, using married as the reference, bachelor's degree as the contrast, and excluding categories with few observations.

Compute the weight of evidence for grade and work sector using log ratios, select higher-woe categories for regression, and address a data entry error in the dataset.

Analyze the distribution of installments with a histogram, assess the cumulative probability, and apply fine classing by dividing observations into 20% percentile groups to create balanced default and non-default categories.

Combine installment groups into a four-category indicator and create four dummy variables to enable tabulation and weight of evidence.

Learn to compute weight of evidence for loan installments using association tables, comparing dummy installments with default indicators, and select regression variables from a balanced dataset.

Generate dummy variables for gender and relationship status in STATA, coding male as 1 and encoding four statuses (divorced, married, single, widowed) before machine learning or logistic regression analysis.

Generate dummy variables for work sector and education level, tabulate to create labeled dummies with reference categories, and set a bachelor's degree indicator to 1 for university degree.

Split the data into 80 percent training and 20 percent testing to train the model on the 80 percent and validate on unseen data, using a random uniform assignment.

Run logistic regression (logit) with categorical predictors, including default, instalments, maturity, and sector, and train on the estimation sample to observe 3,903 observations and interpret results.