Transforming Data to Make Better Predictions
Statistics, Predictive Modeling and Data Mining
Learn to understand the value and pitfalls of transforming data, and how to choose an appropriate transformation that will yield a logical and useful model. Understand how to handle one of the statistical assumptions for regression - that the error (variance) is distributed normally and uniformly across the range of the data. See how to transform data to a new scale to make the error better match this criterion to avoid possibly creating a model yielding physically impossible and potentially embarrassing results, such as negative values of hardness, resistivity, or the number of defects.
This webinar covers: an overview of principles of data transformation, descriptions of situations where transformation is important, and several case studies using Fit Model and Box-Cox transformations.