Create Analysis Notebook 3

[Kibrael]

Create example code and instructions to segment single-family products with the below filters:

• single-family
• first-lien
• owner-occupied
• conventional
• home purchase

This code should be provided as a function that accepts:

• extensions to the WHERE clause (for example: geography, action type, lender)
• table name
• database name
• schema name
• host name

These filters should accept single inputs, or list-like inputs.

This function should have an option that allows the user to write the query results to a pipe-delimited file with a .txt extension.

In the instructions inside the Jupyter notebook, discuss:

• what these filters mean and how they affect the mortgage product
• why a homogenous product is important to analysis
• the presence of action type in the HMDA data and how that affects analysis

Produce the following outputs:

• flat file with a pipe-delimiter and .TXT extension
• Pandas dataframe (shown inline)
• SQL script (located in the SQL folder)
• analysis of a subset of HMDA data showing comparisons of product types in two different states over time. The comparison should use 2004-2017 data that was written to a file and reloaded. This analysis should account for action taken type and use Pandas to generate an aggregate measure of the data.
• one or more example of visualizations of the data. For example originated loan amount averages for several MSAs from 2004-2017.

The goal of this example is to demonstrate how to get a dataset of a homogenous mortgage product, save the dataset to disk, load the data to Pandas, produce aggregate metrics, and graph them in a meaningful way.

[MichaelFurey17]

This can probably be combined with the issue for creating a function library.