BA240 Team Project: Regression Analysis To Find A Linear Regression Model

BA240 Team Project
Search at least THREE Quantitative Variables: One dependent variable and at least Two independent variable. Each variable should contain more than 50 data in order to satisfy the Normality assumption. Use Regression Analysis to find a linear regression model that can be best describes the relationship between the variables you chose.
What kind of variables usually work:
• Factors (e.g., Total Cholesterol Level vs. age and weight )
What kind of variables usually don't work:
• Qualitative Data: (i.e. Location, Color, Model)
• Mutually Exclusive Variables
• Partial vs. Total (i.e., Perimeter = 2*L +2*W, so you can't use Perimeter as a dependent and L and W and independent)
Some Examples:
• Total Cholesterol Level vs. Age and Weight
• Field goals percentage vs. Total years played and Average minutes on play per game
• Homicide rate per 100,000 population vs. Population size, % of families with yearly incomes <$5,000, and Unemployment rate
You may not choose the following data sets:
1) Car price vs. engine size, horse power, mileage, weight, etc
2) Housing price vs. square footage, age, #bedrooms, etc
3) Low Birth Weight vs. Mother's Age and Last menstrual period weight
4) A dataset chosen by another group. First group to get dataset approved gets exclusive use
5) Any other data set used as an example
Tips and Notes
1) When you are doing INDIVIDUAL variable summary statistics (in Part II Data), DO NOT delete "outliers", you need to keep every single value (unless it's missing). For Scatter Plot, you also need to use the original data.
2) Scatter Plot is between TWO variables ONLY (Specifically, Y Vs. X). So, you need to use your dependent variable Y against EACH independent variable. That is, how many independent variables do you have, how many scatter plots you need to produce. Remember to add trend line and show the equation and R-square on the scatter plot. You can simply calculate the correlation using R-square, remember to check the SIGN.
3) Parameters are the y-intercept and slopes, so you need to find out these estimates (b0, b1, b2 …) from the output, and then form your regression equation (the least square line). When you have multiple independent variables, you need to be careful on the interpretations: that is, predict the slopes for independent variables INDIVIDUALLY, which means ONE at a time. When you predict one of the slopes, remember to FIX the others.
4) Compare your models (original model, model after first outlier removal, and model after second outlier removal) in Part IV (Discussion). Which model is better? Keep in mind: Adjusted R-square is NOT the ONLY factor to make decision. You also need to check whether or not the model and/or parameters are significant (look at the p-values).