Lesson 14: NCAA Men’s Basketball Final Four Predictions with Logistic Regression

Previous | Next

Wed, Mar 30

Review:

• Logistic Regression
• Exam 2 Corrections Due Mon Apr 4.

Presentation:

• No lecture today; class cancelled.
• Please spread the word to fellow students.

Assignment:

• Build a Logistic Regression model to forecast the winner of the Final Four basketball games this Saturday.
• This assignment won’t be easy. You will need to think about how to set up the problem and how to manipulate and organize the data to facilitate regression runs. Then you’ll need to figure out how to interpret your results. Take a shot and do your best.
• Use this data sets for your analysis:
  • 2016 Game Results Data (Sheets)
  • Your dependent (y) variable is Column J labelled “Y Range”
  • You will have to decide on your independent (x) variables
  • You may use additional data if you think it would be useful (e.g., Tournament results, Win %, margin of victory, opponent seed, etc.)
• Submit your work as follows:
  • Print Logistic Regression output for your model(s)
  • Specify your projected winner and estimated win probability for both games.
    • For Example, write something like this:
      • “Oklahoma over Villanova, 57% probability of win.”
      • “North Carolina over Syracuse, 63% probability of win”
  • Be sure your name is on all paper(s) submitted
  • Submit to Kim Wharton (or assistant) upstairs in HSB 200
  • [highlight color=”options: yellow, black”]Due before 5pm on Friday, April 1[/highlight] (this is not an April Fool’s Joke; ask for a time stamp)
    • Bonus 1: Tweet your results to me @justinholman
    • Bonus 2: Create a blog post to share results
  • This assignment will count toward a portion of your Exam 3 grade
• Is this the coolest assignment ever or what?

Lesson 13: Logistic Regression

Previous | Next

Mon, Mar 28

Review:

• Exam 2 Results

Presentation:

• Intro to Logistic Regression
• Problem 14.23 using Textbook data

Assignment:

• Read p. 584-589, Logistic Regression
• Complete Problems 14.24 and 14.26 on p. 590-591

Exam 2 Results

Mon, Mar 27

Exam 2 Results:

Corrections:

I will accept corrections for up to 50% credit back on points lost. Corrections must be submitted at the HSB front desk no later than 9:00 am on Mon, Apr 4. You must provide a cover page with instructions, describing what section(s) of the exam you are correcting and what you’ve done to address mistakes or omissions. If I can’t understand your instructions you will not receive any credit back.

Exam 2 Review

Mon, Mar 14

Review for Exam 2:

• Part 1 Take-Home
• Part 2 In-Class
  • Multiple Regression
  • Interpret Regression Output
    • R-Squared
    • Coefficients > y-hat equation
    • t-stats and p-values
    • F statistic and Significance F
  • Building Multiple Regression Models
    • Multicollinearity
    • Interaction Variables
    • Binary Variables

Lesson 12: Binary Variables

Previous | Next

Wed, Mar 9

Review:

• Multicollinearity
• Interaction Variables
• Exam 2
• Reading assignments

Presentation:

• Binary variables
  • also known as “Dummy” variables
  • also known as “Indicator” variables
  • value must be either “1” or “0”
• Demonstration with Real Estate Sample data 

Assignment:

• Use the textbook data (black_7e_excel_database) to experiment with Binary variables
  • Use the “Consumer Food” worksheet
  • Recode the “Region” and “Location” variables as Binary variables
  • Run regression with “Household Income”, “Non-Mortgage Debt” and the newly created Binary columns as independent (x) variables attempting to predict “Annual Food Spending” (y).
  • Eliminate insignificant variables from the model
  • Find the best combination of variables maximizing R-Squared and the F statistic simultaneously
  • Produce regression output for your final model

Lesson 11: Multicollinearity

Previous | Next

Mon, Mar 7

Review:

• Building Multiple Regression Models
• Exam 2 Instructions

Presentation:

• Real Estate Multiple Regression
• Multicollinearity
• Interaction Variables

Assignment:

• Use the Pueblo 2014 SF RE Sales Data (Sheets)
  • Create a new multiple regression model with an interaction variable
  • Mitigate multicollinearity
• Read p. 582-583 Multicollinearity
• Read p. 558-560 Regression Models with Interaction
• Work on Exam 2 Part 1

Exam 2

[highlight color=”options: yellow, black”]Exam 2 on Wed, Mar 16[/highlight]

Part 1 Take Home Instructions:

Download:

• Pueblo 2014 SF RE Sales Google Sheets (Sheets)
• Pueblo 2014 SF RE Sales Excel (Excel)

The last digit of your PID determines your data set. Filter out all “0” records and copy/paste all “1” records to a new worksheet for analysis.

Analysis:

• Use your assigned data set to build the best possible multiple regression model.
• The model should be able to accurately forecast 2015 single family real estate prices in Pueblo.
• Your final model will be evaluated for accuracy and part of your exam score will be based on accuracy.
• Produce and Print the Regression Output for your final model.

Executive Summary:

• Prepare a 1-page summary of your regression analysis and your final model.

[highlight color=”options: yellow, black”]Bring a print copy of your final model Regression Output and the 1-Page Executive Summary to the in-class exam (Part 2) on Wed Mar 16. [/highlight]

Lesson 10: Building Multiple Regression Models

Previous | Next

Wed, Mar 2

Review:

• Regression Output Interpretation

Presentation:

• Building multiple regression models
  • Demonstrate with black_7e_excel_database (Sheets)
  • Manufacturing
    • Cost of Materials (y)
    • Number of Employees, New Capital Expenditures, Value Added by Manufacturer, End of Year Inventories
  • Financial
    • Average P/E ratio (y)
    • Total Revenues, Total Assets, Return on Equity, Earnings per Share, Average Yield, Dividends per Share
  • International Stock Market
    • Nikkei (y)
    • DJIA, Nasdaq, S&P 500, Hang Seng, FTSE 100, IPC
  • Consumer Food
    • Annual Food Spending (y)
    • Annual Household Income, Non Mortgage Household Debt

Assignment:

• Use the Pueblo 2014 SF RE Sales Data (Sheets) data to build a regression model
  • Predict “Selling Price” (y)
  • Use “Beds”, “Baths” and “Total SqFt” as independent (x) variables

Lesson 9: Regression Output Interpretation

Previous | Next

Mon, Feb 29

Review:

• Multiple Regression

Presentation:

• Regression Output Interpretation
  • From Last Time
    • R-Squared
    • Regression Equation Coefficients
  • New Stuff
    • t-Stat and p-values for Independent Variables significance
    • F statistic and p-value (Significance F) for Overall Model significance
    • ANOVA table Sum of Squares values (SSR, SSE, SSyy)
      • SSR = b1^2*SSxx
      • SSE = Sum of Squares residuals
      • SSyy = Sum of Squares of y

Activity:

• Read Demonstration Problem 13.1 on p. 528-529
• Use the Regression Summary Output table on p. 529 to answer these questions.
  • What is the R-Squared value?
  • How many observations are in the data set being analyzed?
  • Name the variables and their corresponding coefficients.
  • What are the t-stat values for each independent variable?
  • What are the p-values for each independent variable?
  • Which variable is the most significant?
  • What is the F statistic for the model?
  • What is the significance level of the F statistic?
  • What is SSE, the Sum of Squares Error?
  • What is SSyy, the Sum of Squares for y?
  • What is SSR, the Sum of Squares Regression?

Assignment:

• Read p. 541-543 Interpreting Multiple Regression Output
• Download Excel data provided as a companion to the textbook: black_7e_excel_database
• Complete the “Analyzing the Database” Problems 1-4 on p. 549 
• Don’t worry about the discussion questions. Instead, produce the models specified and then answer the same “Activity” questions above for each resulting set of Regression Output.