How an AID analysis is useful

A careful study of the tree diagram will show that the AID procedures used 6 of the 25 available categorical independent variables to split the total sample into eight sub groups, each of which utilized information sources with differing levels of intensity.

The group (A11) using the most information sources (an average of 2.90) can be characterized as high school graduates who initially considered more than on bran, and who expected to buy more than one major appliance. The group (B11) using the fewest information sources (an average of 0.87) can be characterized by a head of household who didn’t finish high school, who was older than 45 years of age, and who considered the purchase of a product costing less than $200. Thus, the researchers used an AID analysis to identify eight groups whose average use of information sources ranged from 0.87 to 2.90.

When the researchers compare those averages with the total average of 1.84 they can begin to identify the characteristics of appliance purchasers who make above average use of advertising and other information sources. This information can help the appliance manufacturer prepare more effective marketing programs.

Many marketing studies are concerned with identifying the characteristics of the heavier users of a product or brand, of the more frequent readers of a newspaper or magazine, or of the individuals with a more favorable attitude toward dining out and going to the theater. These things can be the dependent variables in AID analyses if they are measured as continuous variables. Most studies of this type can be analyzed with AID if the researchers have also collected 10 or more categorical independent variables. When such studies have many independent variables, researchers will find that it is too cumbersome to analyze the data with cross tabulations. Rather, AID can be used instead, and it is a more efficient method of analysis in three ways.

AID will select only those variables that can split the total sample in to two sub-samples in such a way that there is a statistically significant difference between the sub-sample means on the dependent variable of interest.

AID identifies the independent variables in the order of their ability to select groups with the largest differences – the first independent variable creates two groups with the largest possible significant differences, the second independent variable creates two new sub-groups that again have the largest possible difference and so on.

AID can help researchers understand the effect on the dependent variable of combinations of two or more independent variables. For example, that being a high school graduate (subgroup A – uses an average of 2.13 information sources) becomes much more important if the individuals also is considering more than one band and is expecting to buy more than one major appliance (sub group A11 – uses an average of 2.90 information sources).

Problems in using AID:

Researchers can encounter a serious problem when using AID if they include in their analysis certain independent variables that are not logically related to the dependent variable. It is worth remembering that AID is a stepwise analysis and that, if the first split is a spurious one, the remaining splits on the tree diagram will probably reflect useless findings. This suggests that AID should not be used as a “fishing expeditions” – analyzing all possible independent variables and hoping for “good” results. Rather the independent variables used in the analysis should be carefully selected.

Another problem with AID is that researchers can’t tell how well the tree diagram fits the data. Unlike regression analysts which calculates an R2 value or LDA which provides a “Range of Misclassification”. The AID procedures are unable to measure how good the results are. There are no statistical or other tests which can help the researchers with this problem.