You are on page 1of 2

Which model fits the data better?

1. The same y
2
Different number of Xs => Based on adjusted – R

Example:

rdintens = β0 + β1log(sales) +u R2=0.061, 𝑅̅2=0.03

rdintens = β0 + β1sales + β2sales2 +u R2=0.148, 𝑅̅2=0.09

 Model 2 fits the data better because the adj R-square (𝑅̅2) is higher

2. The same y
2
The same number of Xs => Based on R

Example:

rdintens = β0 + β1log(sales) + u R2=0.061, 𝑅̅2=0.03

2 2
rdintens = β0 + β1sales + u R =0.148, 𝑅̅ =0.09

2
 Model 2 fits the data better because the R-square (R ) is higher
2
3. Different forms of y (ex: log y and y) => Find the new R for the log model, and then compare R2
2 ̂)
The same Xs ̂ 𝟐 /𝟐)*exp(𝒍𝒐𝒈𝒚
[corr(y,𝒚̂ )] while 𝒚̂ = exp(𝝈

Example:

Open CEOSAL1.DTA

Stata command:

reg lsalary lsales roe

predict lnsalaryhat, xb

gen salaryhat2=exp(.48224/2)*exp( lnsalaryhat)

corr salaryhat2 salary


gen r_square=0.2210^2

display r_square=0.2210^2

Answer:

Model 2: R2= 0.048841

 Hence, the model 2 fits the data better because the R-square (R2) is higher

You might also like