Data, often large amounts from multiple sources, must be collected, cleaned, organized and evaluated.
Was the research designed properly? Were scientific methods properly applied? Are the findings based on correct assumptions? Are there inaccuracies, missing values, or gaps in the logic?
Click on the arrows below to learn more about each service.
Data analysis can discover hidden patterns and suggest hypotheses to test. Statistical models fitted to data can assist in estimating and predicting outcomes. Examples include determining dollar amounts of fraud and theft, and forecasting product lifetimes and failures.
Time series methods in statistics exploit relationships and patterns in data collected over time – such as sales volumes or measures of product quality – and provide a basis for predicting future values.
In many cases, graphics are the most effective way to present and explain data, theories and conclusions. A single clear graphic has helped settle disputes.
Machine learning is a collection of methods to classify observations or predict new ones. Typically these methods work in an automatic way to “learn” from patterns in data.
Quality control includesl statistical techniques that have been highly successful in assuring the quality of products and services. Examples include selecting sampling and analytical techniques for environmental site investigations and evaluating the accuracy and completeness of databases.
Regression analysis includes techniques for modeling multiple factors to predict outcomes such as sales, accidents, market prices or salary. It’s also used, along with other analysis, to infer causal relationships among variables.
Which data and statistical methods will produce the needed answers in the most efficient way? Research design—broadly conceived—answers those questions.
Risk analysis studies decision-making when costs and benefits of different actions are uncertain. Risk analysis has been applied at A&I to assess environmental risks from contaminated sites, compute the likelihood of catastrophic accidents, and evaluate product reliability.
Sampling transactions, products, persons and processes can produce enormous savings compared to a census. Statistical sampling tells you how many samples you need, how best to collect them, and how accurate the results are likely to be.
Simulation is used to model complex systems, investigate how well alternative statistical methods work in given situations, and find out what can happen when theoretical assumptions do not hold.
Statistical validity—the ability of a result to hold up when challenged—is a goal of any analysis. Statistical significance is an objective when estimates or decisions must meet accuracy and precision thresholds.