Hello Applied Stats Community,
Please join us tomorrow, September 16th when we are excited to have
Ben Goodrich (Government/Social Policy) presenting "Bringing
Rank-Minimization Back In: An Estimator of the Number of Inputs to a
Data-Generating Process." The workshop will start with a light lunch
at 12 noon and the presentation will start at 12:15.
You will find a copy of the paper attached and Ben has provided the
following abstract:
This paper derives and implements an algorithm to infer the number of
inputs to a data-generating process from the outputs. Previous working
dating back to the 1930s proves that this inference can be made in
theory, but the practical difficulties have been too daunting to
overcome. These obstacles can be avoided by looking at the problem
from a different perspective, utilizing some insights from the study
of economic inequality, and relying on modern computer technology.
Now that there is a computational algorithm that can estimate the
number of variables that generated observed outcomes, the scope for
applications is quite large. Examples are given showing its use for
evaluating the reliability of measures of theoretical concepts,
empirically testing formal models, verifying whether there is an
omitted variable in a regression, checking whether proposed
explanatory variables are measured without error, evaluating the
completeness of multiple imputation models for missing data, and
facilitating the construction of matched pairs in randomized
experiments. The algorithm is used to test the main hypothesis in
Esping-Andersen (1990), which has been influential in the political
economy literature, namely that various welfare-state outcomes are a
function of only three underlying variables.
We hope you can make it.
Best regards,
matt.
~~~~~~~~~~~
Matthew Blackwell
PhD Candidate
Institute for Quantitative Social Science
Department of Government
Harvard University
email: mblackwell(a)iq.harvard.edu
url:
http://people.fas.harvard.edu/~blackwel/