The astute marketing researcher knows that different marketing questions often require different types of analytical tools. With the wide array of tools at their disposal, researchers need to understand the strengths and weaknesses of each. Here are 11 techniques that researchers may consider—and when. 

1. Multiple Regression Analysis—analyzes one metric dependent variable in relation to multiple metric independent variables. Normality, linearity, and equal variance assumptions must be carefully observed. Often used in forecasting.

2. Logistic Regression Analysis—consists of a variation of multiple regression designed to create a probabilistic assessment of a binary choice, comparing predicted and observed events. Estimates the likelihood of belonging to each group and predicts consumer behavior in the presence of alternate choices.

3. Discriminant Analysis—organizes people and observations into homogeneous groups. Requires metric independent variables with high normality. Used for classifying people according to specific characteristic, for instance, buyer and nonbuyer. Comparable to logistic regression when there are only two categories in the dependent variable, but logistic regression has fewer assumptions to be met.

4. Multivariate Analysis of Variance (MANOVA)—compares categorical independent variables with multiple metric dependent variables, determining dependence relationships across various groups. Used in experiment design. If sample sizes are too large, the model becomes impractical.

5. Factor Analysis—reduces a wide range of related variables into a smaller set of uncorrelated factors. No dependent variables. Major analysis methods include common factor analysis and principal component analysis.

6. Cluster Analysis—creates subgroups out of similar individuals or objects within a large data set. Not all individuals or objects may fit within the defined subgroups. Major clustering methods include hierarchical, nonhierarchical, and a combination of the two.

7. Multidimensional Scaling (MDS)—uses perceptual mapping to convert consumer judgments into multidimensional distances. An exploratory technique, useful when dealing with unknown comparison bases. Typically requires four times as many objects as dimensions.

8. Correspondence Analysis—creates a perceptual map of object attribute ratings through dimensional reduction.
Useful when assessing a wide range of attributes since it does not require that all individuals evaluate every attribute. A combination of independent and dependent variables can result in difficulties in interpretation.

9. Conjoint Analysis—also known as tradeoff analysis, measures the relative value of the features that comprise a product or service. Results can be used to predict product preferences and forecast demand under a wide variety of scenarios.

10. Canonical Correlation—correlates independent and dependent variables simultaneously, using metric independent variables or nonmetric categorical variables. Has fewest restrictions of any multivariate technique, but also fewest assumptions.

11. Structural Equation Modeling (SEM)—employs a variety of techniques (e.g., LISREL, latent variable analysis, confirmatory factor analysis) to simultaneously examine numerous relationships between variable sets. SEM is used to develop or validate theoretical models regarding causal relationships between independent variables and one or more dependent variables.

Questions

1. What questions do you think you should be asking when choosing a multivariate analysis technique?

2. Have you ever done a project in which the analytical technique chosen was not the best fit? What difficulties did you encounter as a result?