Linkage analysis and driver analysis

linkage One of the more interesting tools is linkage analysis. One way to describe linkage analysis is through examples:

Issues which seem to predict real-life retention are statistically identified through multiple regression.
An employee survey from 2020 is combined with customer survey data from 2021 to find the employee survey items which predict customer loyalty or wallet-share.
Using an enrollment database and student survey data, analysts predict retention and academic achievement based on student satisfaction, SAT scores, high school GPA, age, and other factors.

There are many important issues to consider with regard to linkage analysis. First, people use cross-sectional data—for example, we look at different locations at the same time. It is often better to have the predictors from one time and the outcomes from a later time, but this is not always possible. Regardless, we are making many assumptions about causality, since correlations and regressions cannot really show cause and effect—only whether, as one thing goes up, another one goes up (or down).

While the methods are basically the same, most people use the term “driver analysis” for trying to find causes (“drivers”) of an outcome within a single survey or other dataset; while “linkage analysis” uses more than one data source, linking them together.

Some of the methods used are:

MULTIPLE REGRESSION -Regression is the most accessible method, as well as one of the most robust. In most cases, regression will be enough to show a relationship.
DISCRIMINANT ANALYSIS - This tends to have similar results as regression, but is used to "work backwards."
LISREL - essentially, models are statistically tested to see how well they fit the data. One could generate models using theory or other statistical techniques and check them for goodness-of-fit using LISREL. This tests the entire model rather than a single relationship. (A model could be as simple as employee satisfaction increases customer satisfaction which increases revenue).
CHAID, CART, and other tree-building systems - using a wide variety of techniques, the data are analyzed to see which splits can best predict the outcomes. For example, the decision to buy a particular piece of software can best be predicted first by a customer's prior purchase, then (for purchasers) by their satisfaction and (for nonpurchasers) by their position in an organization.

Often, corporations and other organizations use the "linkage map" to focus their actions on issues which will have the highest impact on some outcome - customer service, profit, turnover, etc. If this is the end goal, then most methods, which are concerned with the "truth" (or goodness of fit) between the whole model and the data, may not be appropriate, because you really want to know what you can do to get the most change in the outcome. Sometimes, it pays to ignore the issue of shared variance and work on the major contributors to an outcome individually.