Automatic Interaction Detector (AID)

Typical Problem Studied with AID: Automatic interaction detector (AID) is typically used when researchers wish to identify the respondent characteristics uniquely associated with each of several different consumption rate categories, such as very heavy consumers, heavy consumers, moderate consumers, light consumers, and non-consumers. That is, AID can be used to describe how people who used four or more information sources before buying an appliance differed from people who used two or three information sources when buying an appliance, and how both of those groups different from people who used fewer than two information sources when buying an appliance. AID can also be used in any study designed to identify the characteristics of heavy users of a product or services and how they differ from the characteristics of moderate users, light users, and nonusers of the product or service.

Companies are often interested in such information because, for many products or services, it is quite possible that as few as 20-25 percent of the population may account for 50-60 percent (or more) of total consumption. When this is the case – that is, when fewer than half of the population account for a majority of a product’s or service’s consumption it is very important for the supplying companies to know a great deal about the characteristics of heavy and moderate users and how they differ from the characteristics of light and nonusers. Such information can be very useful in more effectively directing marketing programs to specific target markets.

In these types of studies it is common for researchers to measure a large number of variables that describe the respondents (e.g. age, income, and so on). Because of the large number of these variables, it is impractical for researchers to attempt to analyze the data using cross tabulation. Researchers will find it more efficient to use AID instead of cross tabulation because AID assists the researchers by systematically searching through the many descriptive variables measured (age, income, and the like) and selecting only the ones that are important.

What AID Does: AID is a stepwise searching procedure that uses one way analysis of variance to split a sample into two sub-samples in such a way (1) that the difference between the average consumption or usage rates of the two sub-example is the largest possible differences that can be found, and (2) that the difference is statistically significant.

The AID Procedures are then applied separately to the two sub-samples. This has the effect of splitting the two sub-samples into four groups, each of which has a different average usage rate when compared with the other three groups.

The AID procedures can then be applied once again to each of the four groups. This will create a total of eight groups, each of which will have a different average usage rate when compared with the other seven groups. Thus, there will be eight groups of respondents whose average usage rates range from very much above the average for the total sample to very much below the average for the total sample. The descriptive variables collected during the study are then used to identify the respondent characteristics uniquely associated with each of the eight groups.

Types of Variables used in AID: A researchers can use AID if the main variable of interest (e.g. consumption or usage rates) is a continuous variable, and all the other variables are categorical. That is, AID can be applied to a set of data consisting of a continuous dependent variable and a number of categorical independent variables. This is different from the types of variables needed for cross tabulation, regression, or LDA. The reader should not in the last two columns that with the addition of AID, a researcher has available a method of analysis for every possible combination of categorical and continuous dependent and independent variables.

AID explains the variation in the dependent variable: AID is like cross-tabulation, regression, and LDA in that AID is used to understand and explain the variation in dependent variables. However, as noted above, AID is different from those there methods in that it is the only method that can be applied to data consisting of a continuous dependent variable and categorical independent variables.