ISS Political Science Workshop
Methods for Small-Area-Estimation using Non-Representative Samples
The topic is "Methods for Small-Area-Estimation using Non-Representative Samples", and will be taught by Roberto Cerina (Nuffield College, Oxford University).
The workshop is aimed at scholars with a basic understanding of statistical methods, and who are interested in learning how to use non-representative samples in social science research.
This may be of particular interest to those using social media data to estimate public opinion.
Further details about the workshop, including an outline of topics covered and requirements, are appended below.
- Methods for Small-Area-Estimation using Non-Representative Samples
November 23 (Sat), 10:00 - 17:00（The workshop is done.）
- Rm 549, Akamon General Research Building, UTokyo (Hongo)
[Building #37 in this map]
- If you are interested in this workshop, please complete the form below at your earliest convenience.
There is no fee for attending, and both faculty and students are welcome to participate.
Please feel free to forward this to anybody who may be interested.
- We will send further details about the workshop at a later date to those who have signed up.
【 SIGNUP FORM 】
The course `Methods for Small-Area-Estimation using Non-Representative Samples' will cover methods to reliably obtain estimates of quantities of interest (e.g. national support for the government), from samples which are not representative of the area of interest (e.g. Facebook users).
The methods are widely applicable to a number of fields, including political science and epidemiology.
Attending the course should open up avenues for researchers to obtain estimates of interest at different geographical levels, for a fraction of the cost of old-school techniques.
First Morning - 10 to 11
1) Simple-Methods for Simple Random Samples (R-Code examples);
2) What is Area Estimation;
3) What are Non-Representative Samples;
4) Why do we need - and where can we find - Population Targets;
Second Morning - 11:15 to 12:30
5) Regression Methods to smooth survey data (R-Code examples);
6) Post-Stratification to obtain area estimates (R-Code examples);
- Lunch Break - 12:30 to 13:30
First Afternoon - 13:30 to 14:30
7) Cost-comparisons between MRP and SRS;
8) Non-linear models to smooth survey data (R-Code examples);
9) Multiple-Imputation to expand Population Targets (R-Code examples);
Second Afternoon - 14:45 to 16:30
10) Group Exercise
Students should have a basic understanding of statistical methods, though a broad range of levels is welcome; they should know ad-minimum how to run a linear regression on any software, interpret its coefficients and understand the underlying assumptions.
Prior to the course, students should install R (https://www.r-project.org) and RStudio (https://rstudio.com) from the official website; JAGS (http://mcmc-jags.sourceforge.net) is also needed.
With respect to R packages, students should ensure the following packages are installed prior to the workshop: R2Jags; missForest; ranger; missRanger.
Kenneth Mori McElwain firstname.lastname@example.org
Institute of Social Science
University of Tokyo