12.4 Multivariate Analysis Techniques

Multivariate analysis techniques are powerful tools for uncovering complex relationships in data with multiple variables. These methods help marketers segment markets, position products, and understand consumer behavior by examining how various factors interact and influence each other.

Factor analysis and discriminant analysis are key multivariate techniques. Factor analysis identifies underlying factors explaining variability in variables, while discriminant analysis predicts group membership based on predictor variables. Cluster analysis groups objects into clusters based on similarity, aiding in market segmentation and customer profiling.

Multivariate Analysis Techniques

Purpose of multivariate analysis

Top images from around the web for Purpose of multivariate analysis

• 

  Reading: Influences on Consumer Decisions | Introduction to Business View original

  Is this image relevant?

• 

  The Purpose of Market Segmentation and Targeting | Principles of Marketing View original

  Is this image relevant?

• 

  Putting It Together: Consumer Behavior | Principles of Marketing View original

  Is this image relevant?

• 

  Reading: Influences on Consumer Decisions | Introduction to Business View original

  Is this image relevant?

• 

  The Purpose of Market Segmentation and Targeting | Principles of Marketing View original

  Is this image relevant?

1 of 3

Top images from around the web for Purpose of multivariate analysis

• 

  Reading: Influences on Consumer Decisions | Introduction to Business View original

  Is this image relevant?

• 

  The Purpose of Market Segmentation and Targeting | Principles of Marketing View original

  Is this image relevant?

• 

  Putting It Together: Consumer Behavior | Principles of Marketing View original

  Is this image relevant?

• 

  Reading: Influences on Consumer Decisions | Introduction to Business View original

  Is this image relevant?

• 

  The Purpose of Market Segmentation and Targeting | Principles of Marketing View original

  Is this image relevant?

1 of 3

• Analyze data with multiple variables simultaneously to uncover complex relationships
• Examine how multiple variables interact and influence each other
• Provide a comprehensive understanding of the data by considering multiple factors
• Applications in marketing research include:
  • Market segmentation: divide market into distinct groups (demographic, psychographic)
  • Product positioning: determine how products are perceived relative to competitors (perceptual mapping)
  • Consumer behavior analysis: understand how various factors influence purchase decisions (price, brand, features)
  • Brand preference studies: identify factors that contribute to brand loyalty (customer service, product quality)

Concepts of factor analysis

• Data reduction techniques identify underlying factors or components explaining variability in variables
  • Factor analysis assumes latent variables (factors) influence observed variables aims to identify and interpret factors
  • Principal component analysis (PCA) transforms original variables into uncorrelated principal components each a linear combination of original variables
    • First principal component accounts for largest variability in data (explains most information)
    • Subsequent components explain remaining variability in descending order
• Interpretation of factor and component loadings
  • Loadings represent correlation between original variables and factors or components
    • Higher loadings indicate stronger relationship (variable highly influenced by factor or component)
    • Loadings close to zero suggest weak or no relationship
  • Variables with high loadings on same factor or component are related and can be grouped together

Interpretation of discriminant analysis

• Predict group membership based on set of predictor variables by developing discriminant function maximally separating groups
• Interpretation of discriminant function coefficients
  • Coefficients indicate relative importance of each predictor variable in discriminating between groups
    • Larger absolute values contribute more to group separation (more influential in predicting membership)
    • Positive coefficients associated with higher scores on predictor variable for one group negative coefficients for other group
• Assessment of predictive accuracy
  • Classification matrix shows number and percentage of correctly and incorrectly classified cases
    • Diagonal elements represent correct classifications off-diagonal elements misclassifications
  • Hit ratio represents overall accuracy of discriminant function in predicting group membership (percentage of cases correctly classified)
  • Press's Q statistic tests significance of classification accuracy compared to chance (random assignment to groups)

Process of cluster analysis

1. Select variables to be used for clustering (relevant to research objectives)
2. Choose appropriate distance measure to calculate similarity or dissimilarity between objects
   • Euclidean distance: straight-line distance between two points in multi-dimensional space (most common)
   • Manhattan distance: sum of absolute differences between coordinates (city-block distance)
3. Select clustering algorithm to group objects into clusters
   • Hierarchical clustering: builds hierarchy of clusters by repeatedly merging or dividing clusters (agglomerative or divisive)
   • K-means clustering: partitions objects into pre-specified number of clusters based on minimizing within-cluster variation
4. Determine optimal number of clusters using criteria such as:
   • Scree plot: plots number of clusters against within-cluster variation (elbow point suggests optimal number)
   • Silhouette plot: measures how well each object fits into its assigned cluster compared to other clusters (higher average silhouette width indicates better clustering)
5. Interpret and profile resulting clusters
   • Cluster centroids represent average values of variables for each cluster (mean or median)
   • ANOVA and chi-square tests identify variables significantly differing across clusters
   • Cluster profiles describe characteristics of each cluster based on variable means or proportions (demographics, preferences, behaviors)