[SOCSIM] Two questions about Socsim

[Carl MASON]

Hi  Andrea --

I just heard from Tim a minute ago.  I am pleased to meet you.  I presume
that you are using Bruno's account at UC Berkeley in order to do your
socsim runs.  If so,  It would be preferable to give you your own login for
security and administrative reasons.   It's a simple form :
http://lab.demog.berkeley.edu/Docs/statementofcompliance.pdf

On the substantive questions:

I don't have an answer for #2 -- one could perhaps do some sort of baysean
estimation but it would be a significant project.  I think a little trial
and error is probably the most efficient.  But perhaps I do not understand
the question.

For #1,   there is no direct way of making women change group on birth, but
one could specify birth rates as transition probabilities and set group
fertility rates=1 for all ages.  Effectively transitioning from group a to
group b would have the same rate schedule as birth of a particular parity
and birth would just be a formality after the transition.  But then you
would also have to have some transitions based on the mother's age.  That's
a lot of groups.

If it is the case that the effect of maternal age and parity on mortality
are proportional in the hazard of death and can specified parsimoniously

Ideally something like:

child's mortality hazard = h_0(age)*e^(maternalAge*Beta_1 +
birthOrder*Beta_2)

where h_0 is the mortality hazard rate of a first born  to mothers of some
reference age.

Then it would be pretty easy for me to create a special version of  Socsim
that would assign each person at birth a "death multiplier"
and every time a waiting time to death was generated that multiplier would
shift the specified mortality rate.

You would be able to specify the Beta parameters in the .sup file.


If you wanted to handle maternal age in a fancier way, you could still have
mothers change groups at certain ages and if the child inherits group from
mother --that would determine a particular mortality schedule -- which
would then be shifted by the birth order effect.

This could be kind of fun.

Please send me the form as soon as you get a chance so we can get your
account set up.

Also, there is a low traffic socsim email list that you might want to join:
http://lists.demog.berkeley.edu/mailman/listinfo/socsim

I will cc that list in case some of the other Socsim experts have some
better ideas particularly about the stability question.

thanks.

--Carl




On Thu, Jan 30, 2014 at 8:56 AM, Andrea Verhulst <
andrea.verhulst at uclouvain.be> wrote:

> Dear Carl,
>
> I'm a Belgian PhD student from UCLouvain working with Socsim for my
> dissertation (Bruno Masquelier is my tutor).
>
> Here is my application: http://paa2014.princeton.edu/abstracts/142891
>
>
> I would have two questions if you find the time to answer them:
>
> 1) I need to introduce different child mortality rates depending on the
> age of mother and the birth order. In the first case, I use the grouping
> function (the mother transmits a group membership depending on her age).
> But, in the second case, I don't find a way to introduce differential
> mortality by birth order. Could women change group each time they give
> birth ?
>
> 2) When I work with very low fertility/mortality levels, my estimates are
> quite unstable if the population size is not very big. Is there a way to
> calculate the necessary population size for each demographic regime in
> order to have a similar dispersion in all my simulations ?
>
> Best Regards,
>
> Andrea Verhulst
>
> P.S.: right now I’m in Chile at CELADE, and Tim Miller asked me to greet
> you !
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.demog.berkeley.edu/pipermail/socsim/attachments/20140130/b84b2e08/attachment.htm