[SOCSIM] New version of Socsim ready

[Carl MASON]

The version of Socsim that runs on the demography lab system will be
updated this afternoon around 4PM PST --  as a result, Socsim will be
unavailable for approximately 8 seconds.  It might make sense for some
people to continue with the existing version, if for example, you happen to
be in the middle of a series of runs.  To do so, you will need to call
"socsim.old" where you presently call "socsim".

The new version has two improvements:

1) a new algorithm for modifying fertility rates to account for birth
intervals.  The new routine treats births as a non homogenous poisson
process and adjusts by estimating the probability that a birth will have
occurred within bint months.  The old version used an adjustment based on
the renewal-reward theorem and relied on stronger and not quite exactly
true assumptions.

I expect to improve the routine for parity specific fertility rates in the
near future, but the new version is already considerably better than the
old even in that case.  Be aware that if your fertility rates vary by
groups AND females are subject to group transitions during their ages of
fertility, the bint adjustments will not be right and rate tuning will
continue to be necessary.

2) Internal changes to some sorting routines have greatly increased the
speed in situations where the marriage queues are long.  For example in
simulations that use the "one queue" marriage market and where female
marriage rates are low, a large proportion of the living male population
will be on the male marriage queue.   The new version of Socsim speeds
these simulations up considerably.

For most simulations, these changes should NOT result in any significant
changes in results.  HOWEVER,  running a simulation under the new version
will produce a different population instance from that of the old (current
version) even if the same random number seed is used.

Further, if you are using carefully tuned (high) fertility rates with long
birth intervals, the new version might screw things up for you.  If you
have not invested countless hours in tuning your fertility rates, then
 please use the new version and help me discover the bugs.

Thanks.

--Carl
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.demog.berkeley.edu/pipermail/socsim/attachments/20140113/df2edb5f/attachment.htm