Assignment 3

Minicase – Part II

Make sure you watch the video fully in the Assignment folder before you read this: Minicase – Part II and address the Tasks outlined herein

Background

We continue our minicase on BostonRealty (to help with perfect recall of Part I, I have included a recap at the end of this document).

In your role as the Business Development Manager (BDM), you have been working with your data science team (DST) and have now

1) Identified anomalies and resolved them

2) Identified variables whose associated coefficient estimates have signs that seem to run counter to your intuition and to theory on how they should impact assessed values.

Going forward, you, as the BDM, want to work with the DST to develop a Final Cut Predictive Model by drawing on what you have learned and the insights you have gained from the content in BAN 601 Modules 1, 2, 3a, and 3b. The main guidance you, as the BDM, give the DST is that you want a model that balances predictive power with explanatory power. So, for instance, you don’t only want a model that minimizes the RMSE. You are willing to trade RMSE for explanatory power (say, measured by the adjusted R-squared) – you want a low RMSE but you also want a model with a high adjusted R-squared and, specifically, one where all the signs of the coefficient estimates make intuitive and theoretical sense. This does not mean that all the signs have to be positive or all the signs have to be negative.

Another piece of advice you offer is that in the design of the model, it is not important that all the variables that were included in the First Cut Predictive Model should appear in the Final Cut Predictive Model. You must think about whether the main information content of a variable is contained in another variable or whether two variables can be combined into one (so you have higher degrees of freedom and enhance your adjusted R-squared). Please note, though, that you have to have a good rationale to drop a variable or to combine variables.

Task

Your task, as a member of the data science team, is to build a Final Cut Predictive Model based on the guidance offered by the BDM. You are to report your results in a Microsoft Word document in the multipart Table given below.