Re: Linear Model with Missing Data Values

Frank Simon wrote:
> Hello,
>
> I am a student from germany and I am going to write my "Diplomarbeit"
> about linear model theory.
>
> y=X*b+e e~N.V.(0, S)
>
> The problem I have to deal with is, is to handle missing values in the X
> matrix. There is many literature about solving this problem and there
> are very simple solutions like hot deck imputation or cold deck
> imputation, which I don't find very satisfactory for a "Diplomarbeit".

Perhaps you need to explain briefly the reason why. relative to the
"Diplomarbeit", because I think you'll find those and related terms
in the subject.

> I only found one book ! Does anyone has this book?

I am sure Rod and Don do. :-) I don't.

> Little, Roderick J. A.; Rubin, Donald B.
> Statistical analysis with missing data

Art Kendall gave a reference for a recent edition.

> And it is worth buying?

For those seriously interested in the subject. Rod Little and I
were multi-term Assoc Editors for JASA Applications in the late 1970
- 1980s era, while Don Rubin was Editor in one of those terms.
Almost everything they wrote are worth reading, IMO. Whether a
book is worth buying is a slightly different matter. I am a
cheapskate when it comes to buying books, because when I needed
them, they were usually found in libraries.

Don has been using missing data techniques for decades in problems
related to the US Census, funded by our Uncle Sam. So, at least
Uncle Sam would recommend him.

-- Bob.

.