Let the model be: Yi B + BTi + i=1,...,n but suppose you are missing data...

 

Transcribed Image Text:

Let the model be: Yi B₁ + B₂Ti +€į i=1,...,n but suppose you are missing data on the nth observation of r. To deal with the missing data, you do mean imputation, setting Tn = Tn-1 = (n − 1)-¹₁, then estimating 3₁ and 3₂ based on all n observations. (a) Show that, as n→∞, this procedure gives you the same coefficient estimates as simply estimating ₁ and 32 from the n-1 observations for which you have complete data. Solution (b) Compare the R-squares from the two regressions. Does goodness-of-fit depend on the approach you take to the missing data? Let the model be: Yi B₁ + B₂Ti +€į i=1,...,n but suppose you are missing data on the nth observation of r. To deal with the missing data, you do mean imputation, setting Tn = Tn-1 = (n − 1)-¹₁, then estimating 3₁ and 3₂ based on all n observations. (a) Show that, as n→∞, this procedure gives you the same coefficient estimates as simply estimating ₁ and 32 from the n-1 observations for which you have complete data. Solution (b) Compare the R-squares from the two regressions. Does goodness-of-fit depend on the approach you take to the missing data?

Answer rating: 100% (QA)

a We have y1 1 2x1 1 y2 1 2x2 2 yn 1 2xn n where 1 2 n are iid er... View the full answer