statistical techniques and SD

[duilio]

Hi all,
is there any good book you can recommend about statistical techniques that can be used in SD (for parameter estimations mostly) ?

Or, put another way, what are the techniques that you used most as modellers?

thanks in advance!!
Giovanni

[duilio]

Hi all, I probably asked a too-wide question. The question can be re-forlulated in these terms, according to your experience as modelers what are the statistical techniques that you applied more frequently ? Regressions, time-series modelling (even though I don't think this one will be very useful)...

If it's still too-wide no worries!
G

[LAUJJL]

Hi Giovanni

I think that the most used technique is regression or sampling techiques.
But you may find the same results using different statistical techniques. I think that before wanting to estimate parameters using for instance sampling techniques, it is better to use sensibility analysis to estimate the level of accuracy necessary to the parameter estimation to avoid wasting time. I am generally very sceptic about parameters estimation and prefer to build robust polices that work within a wide range of parameters confidence interval, instead of wasting time trying to estimate accurately the values of parameters and using ellaborate methods like Anova etc, to analyze the fiability of average estimation for example.
Regards.
JJ

[duilio]

Thanks JJ, your comments are always valuable.

G

[tomfid]

Appropriate statistics for estimating dynamic models are described in Fred Schweppe's Uncertain Dynamic Systems, which is unfortunately out of print.

David Peterson described the approach for SD in his thesis, http://dspace.mit.edu/handle/1721.1/27424 , which is more or less the basis of the implementation of Kalman filtering and optimization in Vensim.

I discuss a little bit in a series of blog posts, starting here

http://blog.metasd.com/2009/02/sea-level-rise-models-i/