Email discussion lists for the UK Education and Research communities

## SIMSOC@JISCMAIL.AC.UK

#### View:

 Message: [ First | Previous | Next | Last ] By Topic: [ First | Previous | Next | Last ] By Author: [ First | Previous | Next | Last ] Font: Proportional Font
 LISTSERV Archives SIMSOC Home SIMSOC 2001

#### Options

Subject:

Re: Algorithms for Simulated Populations with Biographies

From:

Date:

Thu, 3 May 2001 14:13:31 +0100

Content-Type:

text/plain

Parts/Attachments:

 text/plain (53 lines)
 Edmund, would I be right that this sort of simulation would then pass on the class of the child to that of its children (according to some probability function) and so on ad infinitum? If so then is this not a Markov chain and you would expect it to converge to a stable distibution regardless of the initial distribution? If so what you need is some way of measuring the convergence. But first you need to decide that it is Markov and has no meaningful feedback.... Alan > -----Original Message----- > From: News and discussion about computer simulation in the social > sciences [mailto:[log in to unmask]]On Behalf Of Edmund Chattoe > Sent: 03 May 2001 12:42 > To: [log in to unmask] > Subject: Algorithms for Simulated Populations with Biographies > > > Dear All, > > I am trying to generate a simulated population with a particular > distribution attributes. The snag is that some of these attributes > have a "biographical" or historical dimension ie the social class of > a child is the social class of its father - I know, sexist, but the > research norm _when the child is born_. This involves some "children" > with "fictional" parents at t=0 but these fictional parents also need > to have sensible values of the biographical attributes. It is quite > tricky to "match" initial attribute distributions with those that > then emerge in the evolution of the population. Either it takes > forever for any initial distortion to "work out" or it never does and > the population is unstable. > > Has anyone done something like this? Is there an algorithmic "trick" > I'm missing? I'd like to be able to use biggish populations (1000+) > so any solution can't be too computationally lazy. > > ATB, > > Edmund > -- > ========================================================================= > Edmund Chattoe: Department of Sociology, University of Oxford, 3 George > Street Mews, Oxford, Oxon, OX1 2AA, tel: 01865-278833, fax: 01865-278831, > http://www.sociology.ox.ac.uk, Review Editor, J. Artificial Societies > and Social Simulation (JASSS) http://www.soc.surrey.ac.uk/JASSS/, > "So act as > to treat humanity, whether in your own person or in another, always as an > end, and never as only a means." (Immanuel Kant, Fundamental Principles) > ========================================================================== >