[R] Replacing NA values in one column of a data.frame

[Bert Gunter]

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
Behalf Of Steve Lianoglou
Sent: Tuesday, August 18, 2009 7:25 AM
To: John Kane
Cc: r-help at r-project.org; Steve Murray
Subject: Re: [R] Replacing NA values in one column of a data.frame


...** But do NOT do this.***  Unless you are using R to prepare your data
for some other statistical analysis system (most unwise), there is **never**
any reason in R to replace NA's with nonsense numerical codes -- and a great
many reasons NOT to do this, to whit:

1) NA's are a special value in R (actually many, of different types) with
extensive built-in capabilities to handle them properly, deal with them
(often automatically) in fitting functions etc. Recoding to a nonsense
numeric defeats all this careful machinery.

2) You are just begging for trouble by using nonsense numerics: if you
forget your coding or someone else uses your data who's not aware of it,
voilá ! -- you have just guaranteed that data analyses will be partial or
complete garbage.

Finally -- a request to well-meaning helpeRs: Just because you CAN do
something in R statistically or programmatically does not mean you should.
When someone requests something that you believe is the wrong thing to do, I
think it within the bounds of both R etiquette and professional practice to
tell them NOT to do it and explain why not. Of course, sometimes such
admonitions are themselves ill-advised (as this may be), but an open,
courteous, professional exchange on the issue will itself be informative to
useRs.

Cheers to all,

Bert

Bert Gunter
Genentech Nonclinical Biostatisics