Re: st: Count variables and growth curves

[Maarten Buis <maartenlbuis@gmail.com>]

On Thu, Jun 28, 2012 at 12:05 PM, Brendan Churchill wrote:
> I am grouping year of birth into small five-year birth cohorts. I have year of birth from 1909 to 1986. When I recode this into 14 birth cohorts (for example 1981-1986 = 14). These have been coded from 0 to 14. I have also tried 1 to 15, but still the erroneous constant problem remains.

I have two guesses:

1) You made an error when coding this variable and the new variable
does not go from 0-14 or 15, but there are still some outliers
present. This will particularly be important when you incorrectly
added the new cohort variable as a continuous variable. Such
coarsening only makes (some very small amount of) sense when you added
them as a categorical variable (i.e. use the factor variable notation,
see -help fvvarlist-)

2) You correctly added this as a categorical variable, but the
reference category is just too small to obtain reliable estimates of
the expected outcome. You will see that the confidence interval around
the constant will be huge.

I would typically prefer not to coarsen birthyear, but instead use
splines to add a flexible effect of year of birth. See -help
mkspline-.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/