7 - ECMT 1020 Lecture 7 Multiple Regression III We assumed...

Info iconThis preview shows pages 1–16. Sign up to view the full content.

View Full Document Right Arrow Icon
ECMT 1020 Lecture 7 Multiple Regression III
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Some Twists We assumed the relationship between Y and X was linear . In many real contexts this assumption is not valid. Consider the number of errors made by a driver (Y) against the level of alcohol in their blood. It would be likely that as the level of alcohol rises the number of errors would increase disproportionately . Additionally, gains from economies of scale will likely increase at a decreasing rate as output expands This would suggest non-linear relationships .
Background image of page 2
Non-linear Relationships If the true relationship is non-linear using a linear regression model to estimate will lead to errors and violation of a key regression assumption . Transformations of the variables are needed in order to employ the Least Squares technique discussed earlier validly. Most transformations will use a conversion to logarithms of either the Y or X variable or both (other transformations possible) . Consider the error vs alcohol % scatterplot
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Example error 0 50 100 150 200 250 300 350 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 alc% errors
Background image of page 4
Transformation- ln y vs. x log error 0 2 4 6 8 0 0.05 0.1 0.15 log error
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Non Linear (in variables) Relationships Consider the average production costs of a firm with respect to output produced Transformed using Ave. Cost - Quantity Relationship Quantity Ave: Cost $ x e Y 1 0 α α+ = x Y 1 0 ln =
Background image of page 6
Promotion impact on Sales Promotion Sales Non Linear (in variables) Relationships Consider the effect of a promotional campaign on sales Transformed using x Y ln 1 0 α + =
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Transforming non-linear (in-variables) models Transform x or y to create a “linear” relationship Generally, taking ln (natural log) of x or y or both will work Other transformations such as powers (x 2 , x 3 ) square roots (y 0.5 , x 0.5 ) and inverse (1/y,1/x) are also possible
Background image of page 8
Non-linear relationships Y X 1 0 10 1 100 2 1000 3 Y 1000 100 10 1 0 1 2 3
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Transformation log 10 Y 3 2 1 0 1 2 3 X Y X log 10 y 1 0 0 10 1 1 100 2 2 1000 3 3
Background image of page 10
Day after recall DAR vs Number of Ads 0 10 20 30 40 50 60 70 80 90 100 0 2 4 6 8 10 12 Number of Ads DAR
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Regression:Untransformed data Regression Statistics Multiple R 0.892 R Square 0.796 Adjusted R Square 0.767 Standard Error 9.913 Observations 9 ANOVA df SS MS F Significance F Regression 1 2680.287 2680.287 27.273 0.001 Residual 7 687.936 98.277 Total 8 3368.222 Coefficients Standard Error t Stat P-value Intercept 36.731 6.251 5.876 0.001 Ads 5.939 1.137 5.222 0.001
Background image of page 12
Residual Plot - untransformed data Ads Residual Plot -25 -20 -15 -10 -5 0 5 10 0 2 4 6 8 10 12 Ads Residua
Background image of page 13

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Transforming the data X Ln X Y Ads Ln (Ads) DAR 8 2.08 83 2 0.69 47 10 2.30 87 1 0.00 23 3 1.10 63 7 1.95 82 2 0.69 55 5 1.61 74 4 1.39 66 DAR vs ln(Ads) 0 10 20 30 40 50 60 70 80 90 100 0.00 0.50 1.00 1.50 2.00 2.50 ln(Number of Advertisments) DAR
Background image of page 14
Regression: Transformed data Regression Statistics Multiple R 0.980 R Square 0.961 Adjusted R Square 0.955
Background image of page 15

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 16
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 55

7 - ECMT 1020 Lecture 7 Multiple Regression III We assumed...

This preview shows document pages 1 - 16. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online