Re: st: Regression with multiple age groups

[David Hoaglin <dchoaglin@gmail.com>]

Hi, Shirley.

Thanks for that link.  I have not studied the data (for 2010) in
detail, but the data on age at marriage of the husband and the wife
involved in the divorce (or annulment) may give an indication of what
you have to work with.  The table on age of marriage cross-classifies
the 119,589 divorces in England and Wales in 2010 by the age at
marriage of the husband and the age at marriage of the wife, using the
categories that you mentioned earlier. If you have such a crosstab for
each of 20 years, you could analyze the behavior of the marginal
distributions (husband's age, wife's age) over time, and also analyze
the asssociation between those two ordered categorical variables over
time.

Two other tables cross-classify the duration of marriage at divorce by
age of wife (resp. husband) at marriage.  And the data file contains
other tabulations.

It's not clear to me how such crosstabs relate to the main goal of
your project, forecasting divorce rates, but I am a statistician, not
a demographer.  Do demographers (or sociologists) use a standard
definition of "divorce rate"?  Earlier I expressed concern about using
the number of marriages in the same year for the denominator.  If you
had divorce rates (using a standard definition), perhaps disaggregated
by such characteristics as age of husband and age of wife, you might
be able to model the behavior of the characteristics and combine the
results with other variables (such as female unemployment rate and
male unemployment rate, but for the current year or going back a year
or two?) to predict divorce rates.

Apart from any computing, it would be helpful to write out a "model"
that relates divorce rate to the data that you have available, as a
basis for understanding what's feasible.  That model might not be the
same as either of the two forecasting models.  The variation in
divorce rates is probably related to some of the available variables,
but that relation does not necessarily mean that those variables
"explain" the variation in divorce rates.

David Hoaglin

On Fri, Apr 27, 2012 at 10:35 AM, Shirley Sy <shirleysy@hotmail.co.uk> wrote:
> Hi David,
> For individual observations, say for husband's and wife's age, I have the total number of divorces in a particular year where the husband/wife was 'under20', '20to29' etc... and for duration of marriage, each subgroup has the total number of divorces that lasted 'under 2years', '2to5years' etc...
> Just in case my explanation is poor, here's the link to the data that I am using: http://www.ons.gov.uk/ons/publications/re-reference-tables.html?edition=tcm%3A77-238035
> My project is basically to try and explain the variation in divorce rates over time given the available data I have and compare the use of two forecasting models. Apologies for my poor knowledge on regression, I know a lot of theory and very little practical application of it.
> Shirley

*
*   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/