Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: How to correctly code categorical variables effects in FE/RE model


From   Herman Haugland <[email protected]>
To   Statalist - Send questions <[email protected]>
Subject   Re: st: How to correctly code categorical variables effects in FE/RE model
Date   Tue, 20 Aug 2013 15:45:20 +0200

Thank you very much Nick and Maarten!  One related question, how would
you report those year effects? I am asking because I am basing my
analysis on studies that just say: Year effects? Yes/No.

Is that the correct way to do it?  I mean, I code the year effects, I
get the values for my variables of interest, and then:

a) Should I conclude based on the values of those variables of
interest only?  OR
b) Should I say something about the coefficients for the years?

One last newbie question, for the years, does it matter whether I take
the real value of the year (e.g. 1991), or should I code them as
categorical variables? I suppose is the same, since years have a
natural order, but I just want to make sure I am doing it right.

Thanks for your time and consideration.




Med vennlig hilsen / Best regards,

Herman Haugland

________________

Contact Information:

Tel: +47 930 289 69
E-mail: [email protected]
LinkedIn: /in/hermanhaugland


On Tue, Aug 20, 2013 at 3:36 PM, Maarten Buis <[email protected]> wrote:
> On Tue, Aug 20, 2013 at 3:18 PM, Herman Haugland wrote:
>> Suppose that I want to test for the year effects in the regression.
>> I, so far, have been doing the following:
>>
>> xtreg y x i.year, fe
>>
>> But reading here (sorry for citing Wikipedia):
>> http://en.wikipedia.org/wiki/Categorical_variable#Effects_coding
>>
>> It seems I am doing it the wrong way.
>>
>> How would you correctly code categorical variables (year) effects in Stata?
>
> You did it right. Effect coding is just another way of representing
> exactly the same model. Say we have two groups (years), 1 and 2 and
> they are of equal size and year 1 has a value of 3 and year 2 a value
> of 5. We could say the overall mean is 4 and year 1 defiates -1 from
> that mean and year 2 deviates +1 from that mean. That would be effect
> coding. We could also say that year 1 has a value of 3 and year two
> deviates +2 from that value. That would be the default with factor
> variable notation. Notice that these are just different ways of saying
> exactly the same thing.
>
> You could get effect coding with -contrast-, which can sometimes help
> as that comparison fits a particular application better. It is
> certainly not a case of "correct" versus "incorrect" way of doing
> things, as the models are just mathematically equivalent.
>
> -- Maarten
>
> ---------------------------------
> Maarten L. Buis
> WZB
> Reichpietschufer 50
> 10785 Berlin
> Germany
>
> http://www.maartenbuis.nl
> ---------------------------------
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index