There are several questions I want my students to understand and master in this chapter on Linear Regression:
- What are the four assumptions of linear regression (simple linear and multiple)?
- What is meant by dependent and independent variables? (y is dependent, x are independents)
- What is difference between simple linear and multiple linear regressions?
- What is difference between regression model, and estimated regression equation?
- What is a residual? How is it computed? (actual y – estimated y)
- How do you interpret b1 in simple linear regression?
- How do you interpret b1 in multiple linear regressions?
- What happens when p value for f test is lower than alpha i.e. what do you conclude?
- What happens when p value for t test is lower than alpha i.e. what do you conclude?
- What is the difference between coefficient of determination, and coefficient of correlation?
- What does coefficient of determination explain? (in terms of variation)
- What are df associated with f test and t tests for simple linear and multiple linear regressions?
- How to find f test and t test p values?
- How to write estimated regression equation from coefficient output?
- What is adjusted R2?
- How to compute R2? (R2 = SSTR/SST; and also R2 is square of R)
- What are H0/Ha for f test and t tests?
- When do you reject H0 and when do you fail to reject H0 – for f and t tests?
- How are SSTR, SSE and SST related? (SST = SSTR + SSE); Also note: in SPSS ANOVA output associated with Regression>> Regression is same as Treatment and Residual is same as Errors.
- How to find MSTR and MSE from SSTR, SSE and df? (MSTR = SSTR / p and MSE = SSE/(n-p-1), where p=no. of independent variables)
- How to find f test values from MSTR, MSE? (f = MSTR / MSE)
- How to find t test values from b1 and Sb1? (t test value = b1 / sb1)
- When does f = t2 (for simple linear or multiple linear regression)?
- Estimate y for any particular value of x. (Consider the x and y units while estimating the y value).
- Degrees of freedom in terms of n and p, where n= sample size, and p= number of x variables:
- Df numerator (Df1) = p
- Df denominator (Df2) = n-p-1
- Df total (DfT) = (n-1)
- Sample size = Df total + 1