In: Statistics and Probability
Why should we be concerned about using raw versus relative weights when conducting statistical tests in a software package (assuming conventional, non-survey procedures)?
Raw data also known as primary data collected from a source. For example, information entered into a database is often called raw data. The data can either be entered by a user or generated by the computer itself.
Raw data remain available in the foreseeable future for other workers to analyse further
The data can be easily copied into other applications, making analysis by others a practical proposal
The data are available for effective meta-analysis
If raw data is used in statistical package they would remain available. Furthermore, if the data were published within the electronic version of a paper they could not become separated or lost as they would be an integral part of the paper. Also analyst could easily add to or alter the database and rerun the statistical analysis in the knowledge that the analysis would be identical with that performed in the statistical package.
Relative Weights is a way quantify the relative importance of correlated predictor variables in regression analysis. “Relative importance” in this context means the proportion of the variance in y accounted for by xj. Put another way, it helps you figure out what variables contribute the most to r-squared.Many relative importance of indices have been proposed over the years, including the Product Measure. It always produces clear results even if the predictors have very high collinearity.
Reasons to using relative weight
Estimating significance without bootstrapping
You obtain the statistical significance of regression scores by running the method on bootstrapped data. This increases the time needed many times over, than when running a single model.
Computing importance for other types of regression models.