Topic 10- Multiple Regressions Flashcards
Multiple regression model
Shows several independent variables to influence y (dependent variable)
Multiple regression equation
Y=a+b₁x1+b₂x₂+b₃x₃+…+e
b are regression coefficients
A=y intercept
Independent variables are x
Multiple regression test statistic
Ti=bi-β
/
se(b)
~T(n-j)
Same as normal regression test statistic, except the i’s added, and crit value found by (n-j), not (n-2)
Method 2: confidence intervals
Same as regression before, if 0 doesn’t lie within the interval, we can reject null.
Method 3: pvalue
P value less than significance level, we can reject.
Regression hypothesises
Null: no significant effect/influence on y
Alternate: x has a significant influence on y
How do we check the significance of the whole model?
F test from anova.
If F is high, it means at least something in model is significant in helping to predict the dependent variable
Why use multiple regresssions (3)
Isolates individual effect of each variable
Control variables exclude wrong explanations for relationships between the variable of interest and Y.
Ceteris paribus- allows us to hold other things constant while testing for variable of interest
Note: if we find a positive relationship between food expenditure and age. We don’t know why, it could be due to income, family size, and health.
To check if age is actually significant and a relationship exists between it and food expenditure, age should still be significant when the other variables are included