s-news
[Top] [All Lists]

[S] predict.gam() question: summary

To: "'s-news@wubios.wustl.edu'" <s-news@wubios.wustl.edu>
Subject: [S] predict.gam() question: summary
From: "Piotrovskij, Vladimir [JanBe]" <VPIOTROV@janbe.jnj.com>
Date: Tue, 22 Jun 1999 08:09:02 +0200
Sender: owner-s-news@wubios.wustl.edu
Dear S-PLUS users,
This is apparently a FAQ, however, for those interested I summarize two
responses I received.

Laurie Becker forwarded a summary of early discussion on predict.gam()
behaviour which help to decipher the predict.gam error message.

Rachel Fewster wrote:
I think the problem is with  na.action=na.omit in the original gam fit. The
same problem had me stumped for a long time.  ... The quick answer is that
if you remove the NA's from your data before you fit the gam, and miss out
the na.action = na.omit field, I think predict.gam will work properly.

The last advice really helped to solve the problem. Conclusion: if you want
to impute missing values using predict.gam() fit the gam to the data WITHOUT
records that have these missing values. Don't use the na.action = na.omit
option of gam().

Thank you Laurie and Rachel.

Vladimir
vpiotrov@janbe.jnj.com

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news

<Prev in Thread] Current Thread [Next in Thread>
  • [S] predict.gam() question: summary, Piotrovskij, Vladimir [JanBe] <=