When: Wednesday, May 9, 2014, 9:00am-12:00pm and 1:00pm-3:00pm
Taught by: Dr. Richard Wanner, Prairie Regional RDC Academic Director
Sponsored by: The University of Lethbridge Research Data Centre
With the dramatic expansion of statistical modeling techniques in the past decade, formal statistics courses seldom have time to cover many topics that are important to the correct analysis of data in many areas of research. This workshop will provide an overview of three such topics. Bootstrap weights are used in conjunction with case weights in most Statistics Canada data sets, and increasingly in data from other sources. They adjust for complex sample designs to produce correct standard errors for sample statistics. Researchers frequently encounter missing data for large numbers of cases, either because the respondent refused to answer some items or because the interview was truncated. The workshop will cover the most widely used method of imputing missing data, random multiple imputation. Another way in which the results for any form of regression model may be biased is if many cases are excluded from the analysis because they do not have an observed value on the dependent variable. We will examine two methods to address this problem, the Heckman selection model and a more general two-stage model. All examples use Stata version 13.
Topics to be covered
v Bootstrapping: basic concepts
v Bootstrapping standard errors
v Bootstrap weights: basic concepts
v Using bootstrap weights with Statistics Canada data using Stata
v Approaches to imputing missing values
v Assumptions about missing values
v Imputing missing values by random multiple imputation using Stata
v Binary and ordinal logistic regression with random multiple imputation
v Correcting for sample selection bias using the Heckman method
v Correcting for sample selection bias using a two-stage logit model
You may register by sending an email to Tom Perks (firstname.lastname@example.org) containing your name, status (student, faculty, post-doc), and department or faculty. Researchers with current or past RDC contracts will be given priority, though all are welcome. There is no fee for this workshop.