Statistical disclosure control

Inability to access data is the bane of any researcher.  Confidentiality concerns limit the content and form of individual record data (microdata) that can be released to external researchers.  I'm interested in statistical methods to modify the original microdata before dissemination, so we can protect both disclosure risk (for example, identification of units in the study) and preserve the analytic validity of the original data.  Some specific topics:

• empirical assessment of disclosure risk
• synthetic data techniques and building flexible synthesis models 
• applied case studies so we can better understand the relationship between disclosure risk and preserving analytic validity

Missing data techniques

No data set is foolproof from missing values or erroneous values.  In a survey, sometimes questions are skipped.  In a clinical trial, sometimes patients drop out.  If a scientific measurement is recorded with error, then the true value is missing.  Can we just ignore the missing data?  Usually not because the missingness mechanism is related to other variables in our study.   I'm interested in using statistical methods to handle missing data, beyond ignoring the missing values.  

Applied statistics

Some of the best theoretical and methodological research in statistics has come out of solving 'real' problems.  I'm interested in how statistics is used in other disciplines, and whether currently used methods, or assumptions made, need to be challenged.

BActStud/BComm (ANU), FIAA, PhD - Statistics (Harvard University)

• Multiple imputation for missing indigenous status in criminal offence survey data
• Uncongeniality with multiply-imputed synthetic data
• Disclosure risk assessment for linked micro-data
• Disclosure risk assessment for remote access server output