Tutorials
Conjoint analysis is a popular survey tool used in patent infringement and false claims cases. In this R tutorial we provide a soup to nuts demonstration of how to conduct an elementary conjoint analysis. The required consumer preferences or choices are simulated.
2. Logistic vs LPM vs FFT vs OneR: a Fourway Mano a Mano
Interrogating machine learning tools appears to have now been institutionalized by the General Data Protection Regulation ("GDPR") in the European Union. This codifies peoples distrust of black-box algorithms and the associated "trust-me, I am the expert" culture surrounding it. What does this mean? Among other things, this will enhance the appeal of easier to understand algorithms and most-likely place a premium on human-in-the-loop decision-making processes. We offer a tour of the debate pitting the logistic model versus the linear probability model, the Fast & Frugal Tree, and the One Rule model.
3. Is There an Obligation to Scrutinize Our Data?
Is it necessary for forensic economists to examine data disclosed in litigation? If we should chose to do so, there is an array of tests routinely used in forensic accounting premised on Benford's Law. Benford's Law states that the frequency of digits in financial and economic data are not uniformly scattered. Rather, they follow a logarithmic-type distribution. Here I provide an introduction to these tools and illustrate their use.
4. Confidence Intervals via Bootstrap: tools for Medicaid Audits
Data underlying practically all Medicaid Audits entail a population that contains a disproportionately large amount of zeros. A zero is booked when a particular filing is in compliance.
This artifact is known as a zero-inflated population ("ZIP"). A medicare audit sample containing zero-inflated data results in a zero-inflated, highly skewed distribution. This non-conformity with normality assumptions weakens the theoretical statistical rationale underscoring the confidence intervals built into audit procedures. Here we show why this is a problem and how to use the bootstrap to obtain applicable confidence intervals.
Data may be have changed - either deliberately or unintentionally. So it may be productive to run a few tests before starting any analysis. This tutorial on data drift, aka data shift, continues and earlier theme and continues the survey of available tools and tests to appraise whether someone fiddled with the data.
