From 900b84e0f5ae83448d860d5b42a5627fab62905e Mon Sep 17 00:00:00 2001 From: kutea18 Date: Tue, 28 Feb 2017 23:52:31 +0800 Subject: [PATCH] Have question about answers to 1(a ) and 1(d) Your files are a least 2 years old... not sure if can still get response from you. Even you did, I am not sure if using github is the right way for discussion. Anyway, thanks for having such a good notes --- ch2/answers | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/ch2/answers b/ch2/answers index ecbd22b..e6815d2 100644 --- a/ch2/answers +++ b/ch2/answers @@ -1,5 +1,6 @@ 1. (a) better - a more flexible approach will fit the data closer and with the -large sample size a better fit than an inflexible approach would be obtained +large sample size a better fit than an inflexible approach would be obtained. + YX add on: I would agree on "better" due to large sample size. But what is the effect of small number of predictors (b) worse - a flexible method would overfit the small number of observations @@ -7,6 +8,8 @@ large sample size a better fit than an inflexible approach would be obtained better fit (d) worse - flexible methods fit to the noise in the error terms and increase variance + YX add on: I am wondering if error term is very big, would variance/bias still be relevant. After all the MSE is dominated + by error terms already. How can we tell the difference between high flexibility model and low flexibility model? 2. (a) regression. inference. quantitative output of CEO salary based on CEO