[Population Modeling] Introduction

Fri Oct 9 01:52:31 PDT 2015

Dear all, 

Regarding parameter estimation: the highly interesting books on statistical learning by Hastie and Thibshirani. http://www-bcf.usc.edu/~gareth/ISL/ and http://statweb.stanford.edu/~tibs/ElemStatLearn/ (especially the first one I enjoyed). The notion I got from these is that with all methods whether parametric or nonparametric you have to keep an eye on the trade-off between variability (or dependence on this specific set of data/external validity) and model fit/bias. (check the books for the correct terminology and very clear explanation). The more complex your model, the better it may fit the specific data, but it may still do lousy outside these specific data. Also with the nonparametric methods you usually have some tuning parameter to steer this model complexity. So it is a matter of internal and external validity. 

Another issue I would like to hear others about is that in most of my models we are extrapolating one way or the other. E.g. we do have disease data with limited follow-up time and extrapolate prognosis beyond that period. I do not see how to do this, without making assumptions on the functional form, i.e. using some parametric modeling. But would like to hear from others. 

Mostly for the transitions we use individual level data (quite often cohorts, but also GP registry data, with much less certainty on the population (people move) and hence more assumptions needed). For effects of treatments quite often we do use literature/published estimates at aggregate level. I think assumptions are all over, both in the conceptual model, but also when analyzing the data and estimating parameters. Agree that it is very important to document these. (but easy forgotten sometimes)

Regards, Talitha

TL Feenstra, PhD
health economist
RIVM, Centre for Nutrition, Prevention and Health Services Research
Po Box 1
3720 BA Bilthoven
*31 30 274 4387 (mon)
* 31 6 1186 87 60
UMCG, dept of epidemiology
PO Box 30.001
9700 RB Groningen
*31 50 361 5110 (tue/thu/fri)

-----popmodwkgrpimag-news-bounces at simtk.org schreef: -----
Aan: "Hyder, Ayaz" <hyder.22 at osu.edu>
Van: Jacob Barhak 
Verzonden door: popmodwkgrpimag-news-bounces at simtk.org
Datum: 10/09/2015 04:26AM
cc: popmodwkgrpimag-news at simtk.org
Onderwerp: Re: [Population Modeling] Introduction

Welcome Ayaz,
 There are several interesting discussions taking place regarding estimation and assumptions. You can find these in our archives: 
 https://simtk.org/pipermail/popmodwkgrpimag-news/
 What is your approach regarding parameter estimation? Do you rely on human decision making when selecting parameters, or do you strictly use numerical techniques after models structure is determined? 
 Also what kind of modeling assumptions do you usually take? 
 Your infectious disease modeling papers deal with validation. It would be interesting to learn about your modeling assumption process. 
         Jacob

On Oct 8, 2015 8:43 PM, "Hyder, Ayaz" <hyder.22 at osu.edu> wrote:

 Hello Everyone 
I am a new faculty member at the College of Public Health, Ohio State University. I stumbled across this group by accident and I’m very glad that I did. I’ve mentioned a little more about my research below and look forward to fruitful discussions  with fellow modelers. 

My research focuses on understanding the role of multiple determinants of human health (environment, social, biology) operating at multiple levels (individual, neighborhood, regional). I use mathematical and computational approaches (e.g. systems  science methods) to bring together theory, data and methods from multiple disciplines. In the past I’ve worked on a microsimulation model for esophageal cancer and cost-effectiveness analysis (manuscript under review), satellite-based air pollution exposure  assessment and birth outcomes (paper), and predictive validation of agent-based models for influenza (paper 1, paper  2).

 Current areas of research include complex systems models to predict childhood asthma, the integration of agent-based models of residential segregation and coupled economic-disease models of poverty traps, predictive models of infant mortality, and agent-based  models of the food environment in Columbus, Ohio. I am also interested in philosophical discussions surrounding “Big Data” and what we mean by “model”. 

I am currently looking for funding opportunities to fund my research ideas and I am open to collaborations. I would be happy to receive resumes of potential graduate students and postdocs interested in working in my research group. 

Thanks 

Ayaz Hyder, PhD
 Assistant Professor, Division of Environmental Health Sciences 
Affiliated Faculty, Translational Data Analytics
 College of Public Health, The Ohio State University

614-247-4936 | http://cph.osu.edu/people/ahyder   

_______________________________________________
 PopModWkGrpIMAG-news mailing list
 PopModWkGrpIMAG-news at simtk.org
 https://simtk.org/mailman/listinfo/popmodwkgrpimag-news

_______________________________________________
PopModWkGrpIMAG-news mailing list
PopModWkGrpIMAG-news at simtk.org
https://simtk.org/mailman/listinfo/popmodwkgrpimag-news

Proclaimer RIVM http://www.rivm.nl/Proclaimer
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://simtk.org/pipermail/popmodwkgrpimag-news/attachments/20151009/f7e7e1aa/attachment-0001.html>