You are on page 1of 2

Assignment 1:

Roll No: 515

Badrinath Prathi

Q1: State the difference between simple linear regression model (SLRM) and multiple linear
regressions (MLRM) with reference to the case study.

- SLRM: 01 Independent Variable, 01 Dependent Variable


- Two variables. One variable (Dependent / Predicted), the value of which, we like to
predict; We have another variable (Independent / Predictor) that value of which is
used to predict the first variable.
- Linear: Can be expressed in a linear equation
- What is the form?
- Y=Mx+C
- Y: DV
- X: IV
- M: Coefficient of X, the slope of the equation
- C is the intercept
-
- Eg: regress avgcost size
- Dependent variable: Cost
- Independent Variable: Firm size
-
- MLRM: More than One( 1) Independent variable, 01 Dependent Variable
- Can be expressed in Linear Expression:
- Y= M1X1 + M2X2 + M2X3 +…..+ C
- It is Multi -dimensional
- Eg: regress avgcost wage size
- Independent Variables : Wage, Size
- Dependent Variable: Average Cost

Eg:2: regress meanoinstr first year rookie core night day cons
Dependent Variable: meanoinstr (Overall instructor score)
Predictors: first year, rookie, day, nighte.

Here, pair waise predictors like a) first year (-0.232) and rookie (-0.208) b) day (-
0.024) and nighte ( -0.088) have large correlations.
They is no multicollinearity as these variables have precise coefficients.
Q4: How to solve Multicollinearity?

 Coping with MC: Individual Variables:


- Keep all the variables.
- When Sign flipping coefficients on key predictors, it is not recommended to keep
all Independent variables / predictors.
- Omit control variables to reduce omitted variable bias which limits the
interpretation of included variables
- Create a composite score, a single measure which captures the information in
correlated variables.
- Eg: a) An average of two variables that are measured on same scale can be taken.
- B) It helps to avoid problems in interpretation even when you have more
predictors on RHS (Right Hand Side)
- Avoid predictors that are identical.
- Avoid predictors with less theoretical reasoning.
- Try them step by step, one at a time.

You might also like