Class Notes (839,065)
POL222H1 (44)
Lecture 11

# POL222H1 Lecture 11: How to Interpret Linear Regression Results Premium

4 Pages
11 Views

Department
Political Science
Course Code
POL222H1
Professor
Kenichi Ariga

This preview shows page 1. Sign up to view the full 4 pages of the document.
Description
POL222 – Lecture 12 How to interpret Linear Regression Results • How to interpret coefficient • Y = a+BX • B represents how much Y changes, on average, as X changes by one unit • You need to specify a particular unit used for X and Y when you substantively interpret B • Example: Economic voting in US Pres. Elections o Vote = 49.30 +0.75 x Growth o Should may attention to the unit measurement between independent and dependent variable because we need to use it for coefficient o As GDP growth rate improves by 1% point, the incumbent president’s party would gain 0.75 o Depends on unit of dependent variable, so 0.75 represented by vote share • Interpretation could be different depending on the unit of measurement • Question of substantive significance o Is the magnitude substantively large? ▪ We’d like to know if it is politically, economically, or socially important/meaningful o It may not necessarily be clear from the coefficient on X alone o Vote = 49.30 + 0.75 x Growth ▪ It isn’t clear if 0.75 is meaningful ▪ We need a measure • Whether the magnitude of the relationship found between Y and X is substantively large • Whether the magnitude of the relationship politically, economically, or socially important • There is no uniformly accepted term to call this concept • When you interpret or report the linear regression results, you should consider whether the relationship found is substantively significant • However, you need to make an argument whether or not the relationship found is substantively significant • How can we evaluate? • There is no uniform method to evaluate substantive significance • Hence you need to make an argument based on the linear regression results o Your creativity, logical and sensible reasoning, and convincing argumentation are called for o Given the causal theory • Argument based on the coefficient o Sometimes the substantive significance of the relationship may be immediate clear from the coefficient of X—Then you may make an argument based on the coefficient • Political scientist has been interested in the subject because there is a provision of government but that government would not be engage in important political decisions • The upset of the government would be a problem • So what determines the bargaining? • We have not seen the coalition government but in many countries coalition is known and in some governments the duration of bargaining was very long • In many cases substantive significant is not immediately clear • Example 2: standardized coefficient o Compute the change in Y we would expect on average corresponding to a typical change in X observed in data o Compare this expected amount of change in Y to a typical change in Y observed in data o We are going to use a certain typical change to evaluate substantive significance • Standard deviation As typical change o We use standard deviation to represent a typical change in Y and X observed in data o Standard deviation is the measure of variability of a variable • We can see a typical change in variable is much longer in Iran than in Canada • First we would compute the expected change in Y on average corresponding to typical change in Y measured by its standard deviation of the independent variable o B x sd (X) • Then compare this change in Y to a typical change in Y, measured by standard deviation (=sd (Y)) • Application: o Economic voting in the US presidential elections o Y = 49.30 + 0.75 x X • Step 1: compute the change of Y corresponding to a typical change in X, measured by its standard deviation o Standard deviation of the GDP growth rate is 4.32 o B x sd (X) = 0.65 x 4.32 = 3.24 • This number means: one standard deviation change in GDP growth rate would increase vote share by 3.24 • Step 2: compare this change in Y to a typical change in Y measured by its standard deviation o We divided the dependent variable by its standard deviation o Standardized coefficient = B x sd (X)/sd (Y) = 3.24/5.44 =
More Less

Only page 1 are available for preview. Some parts have been intentionally blurred.

Unlock Document

Unlock to view full version

Unlock Document
Me

OR

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Join to view

OR

By registering, I agree to the Terms and Privacy Policies
Just a few more details

So we can recommend you notes for your school.