[R] A regression problem using dummy variables

[Peter Dalgaard]

rlearner309 wrote:
> I think it is zero, because you have lots of zeros there.  It is not like
> continous variables.
>
>   
Think again. The sum of products may be zero, but that is not the 
covariance. And don't dismiss Thomas, he is usually right.

Anyways, the coefs of dummy variables represent differences to the same 
base level, and chosing a poorly determined base level (essentially: 
whose mean is determined by only a few observations) will cause high 
parameter correlation. It should only affect those parameters though, 
and it is not really clear what VIF means for dummy variables. One often 
choses to relevel() to make the largest group the base level, but it 
really comes down to which group contrasts you want to look at.


>
> Thomas Lumley wrote:
>   
>> On Wed, 2 Jul 2008, rlearner309 wrote:
>>
>>     
>>> I think the covariance between dummy variables or between dummy variables
>>> and
>>> intercept should always be zero.  meaning: no sigularity problem??
>>>
>>>       
>> No.  You can easily check that this is not true using the cov() function.
>> Indicator variables for mutually exclusive groups are negatively
>> correlated.
>>
>>      -thomas
>>
>>
>>
>>     
>>> rlearner309 wrote:
>>>       
>>>> This is actually more like a Statistics problem:
>>>> I have a dataset with two dummy variables controlling three levels.  The
>>>> problem is, one level does not have many observations compared with
>>>> other
>>>> two levels (a couple of data points compared with 1000+ points on other
>>>> levels).  When I run the regression, the result is bad.  I have
>>>> unbalanced
>>>> SE and VIF.  Does this kind of problem also belong to "near sigularity"
>>>> problem?  Does it make any difference if I code the level that lacks
>>>> data
>>>> (0,0) in stead of (0,1)?
>>>>
>>>> thanks a lot!
>>>>
>>>>         
>>> --
>>> View this message in context:
>>> http://www.nabble.com/A-regression-problem-using-dummy-variables-tp18214377p18237666.html
>>> Sent from the R help mailing list archive at Nabble.com.
>>>
>>> ______________________________________________
>>> 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.
>>>
>>>       
>> Thomas Lumley			Assoc. Professor, Biostatistics
>> tlumley at u.washington.edu	University of Washington, Seattle
>>
>> ______________________________________________
>> 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.
>>
>>
>>     
>
>   


-- 
   O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark      Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)              FAX: (+45) 35327907