[R] Unexplained behavior of level names when using ordered factors in lm?

[David Winsemius]

On Dec 2, 2011, at 9:51 AM, Tal Galili wrote:

> Hello dear all,
>
> I am unable to understand why when I run the following three lines:
>
> set.seed(4254)
>> a <- data.frame(y = rnorm(40), x=ordered(sample(1:5, 40, T)))
>> summary(lm(y ~ x, a))
>
>
> The output I get includes factor levels which are not relevant to  
> what I am
> actually using:
>
> Call:
>> lm(formula = y ~ x, data = a)
>> Residuals:
>>    Min      1Q  Median      3Q     Max
>> -1.4096 -0.6400 -0.1244  0.5886  2.1891
>> Coefficients:
>>            Estimate Std. Error t value Pr(>|t|)
>> (Intercept) -0.03276    0.15169  -0.216    0.830
>> x.L         -0.28968    0.33866  -0.855    0.398
>> x.Q         -0.38813    0.33851  -1.147    0.259
>> x.C         -0.27183    0.34027  -0.799    0.430
>> x^4          0.25993    0.33935   0.766    0.449

Those are polynomial contrasts: linear, quadratic, cubic and quartic.  
If you don't want contrasts based on ordered factors then just use  
regular factors. You should probably be looking at:

?"C"

(...yet another function whose name should be avoided in naming data- 
objects.)

-- 
David.


>> Residual standard error: 0.9564 on 35 degrees of freedom
>> Multiple R-squared: 0.08571, Adjusted R-squared: -0.01878
>> F-statistic: 0.8202 on 4 and 35 DF,  p-value: 0.5211
>
>
> I am guessing that this is having something to do with the contrast  
> matrix
> that is used, but this is not clear to me.
> Can anyone suggest a good read, or an explanation?
>
> Thanks.
>
>
> ----------------Contact
> Details:-------------------------------------------------------
> Contact me: Tal.Galili at gmail.com |  972-52-7275845
> Read me: www.talgalili.com (Hebrew) | www.biostatistics.co.il  
> (Hebrew) |
> www.r-statistics.com (English)
> ----------------------------------------------------------------------------------------------
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

David Winsemius, MD
West Hartford, CT