[R] NA handling with lm

[Petr Pikal]

Hallo

On 28 Jul 2005 at 18:44, Andreas Cordes wrote:

> Hi,
> I have a problem that is hopefully easily solvable, but I dont find
> the clue in the documentation. I am examining a linear model. One of
> the variables has NA values. Even though na.action=na.omit, i get NA
> as results for this variable. Can I use lm in such a case to get
> estimates? Or do I have to do some form of imputation before doing so?
> Here is the call and the results, hope you can help. Best regards,
> Andreas
> ----------------------------------------------------------------------
> ----------------------- lm(formula = ESSIK ~ ALTER + as.factor(S2) +
> as.factor(S15A) +
>     as.factor(S8) + as.factor(LAND) + as.factor(S18B) +
>     as.factor(BERUF) + as.factor(KIRCHE) + as.factor(H_EINKOM) +
>     as.factor(PARTNERS), na.action = na.omit)
> 
> Residuals:
>      Min       1Q   Median       3Q      Max
> -17.0675  -2.0151   0.4267   2.7644   9.7333
> 
> Coefficients: (2 not defined because of singularities)

Problem is not in NA handling but that some of your coeficients 
can be represented as linear combination of other coeficients. You 
have to omit them.

HTH
Petr


>                       Estimate Std. Error t value Pr(>|t|)   
> (Intercept)          23.755915   1.844110  12.882  < 2e-16 ***
> [...]
> as.factor(BERUF)7    -1.236836   0.701323  -1.764 0.077942 . 
> as.factor(KIRCHE)1   -0.811751   0.237699  -3.415 0.000649 ***
> as.factor(H_EINKOM)2        NA         NA      NA       NA   
> as.factor(H_EINKOM)3        NA         NA      NA       NA   
> as.factor(PARTNERS)1  2.057070   0.342546   6.005 2.23e-09 ***
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
> http://www.R-project.org/posting-guide.html

Petr Pikal
petr.pikal at precheza.cz