Statistical Sampling and Surveys
In many investigations, there’s simply too much information to review. In audits, forensic accounting investigations, review of reinsurance claims, and a variety of other applications, there are tens of thousands or hundreds of thousands of records and files to sift through. The question becomes, how much should we look at?
A sample of records is efficient, cost-effective, can be completed in a reasonable amount of time, and can, in some cases, be more reliable than looking at all of the records or files that are available. Our statistical sampling experts provide sample design, selection, and extrapolation services that can aid any investigation. For the review of large financial portfolios, like securitizations, we have testified as to the use of sampling as a means of accumulating evidence in discovery and the validity of samples for presentation of evidence at trial.
There are many constraints on selecting a representative sample. A sample has to be focused on the right population. Maybe the sample has subgroups to provide information for subgroups. Finally an overly ambitious sample can lead to unnecessarily high costs of data collection. Our samples are designed to estimate multiple values while being cost-efficient.
We use multiple methods and multiple sources to collect and aggregate data. We carry out mail, telephone, and internet surveys, conduct focus groups, and obtain information from observational studies like records reviews in hospitals.
Data enrichment occurs when your sample can be linked to other sources of data, like tax assessment records, Census data, or administrative records describing transactions made by customers.
A sample is only a sample. To be useful, the sample must be extrapolated to the population that the sample represents. We use techniques from sampling theory to obtain the most reliable population value.
Data Modeling and Forecasting
We don’t believe in a one-size-fits-all approach to analysis. We know each client and each project has different needs, different objectives, and different targets. We choose the right tool for the data and the questions being tested.
To compare two choices, we embed an experiment in a survey to test if there is a material difference in the items being compared. These techniques are used to clear away the clutter and obtain a valid comparison.
Questions can get complicated when multiple factors abound and confuse the basic issue being addressed. We use multiple variables simultaneously to control for external factors while testing the key issue under study.
Processes flow over time and the questions being asked may relate to a change or difference that occurs over time. We use standard econometric techniques to understand change over time.
There are times that many factors are changing simultaneously in a complicated process. Monte Carlo simulations are a way to deal with uncertainty in each of these factors, to vary assumptions, and to obtain a clear picture of potential outcomes.