Unlike univariate and bivariate statistical methods, a multivariate analysis can analyze more than one relationship at a time. It is a broad term and there are several multivariate data analysis methods, each with its own purpose. Many of these methods involve analyzing data in depth, across multiple dimensions such as gender, age, or segment. One of the key advantages of multivariate analysis is that it allows patterns and trends within the data to emerge which might not have been clear when only analyzed at a single level. There are many areas in market research that would benefit from multivariate analysis such as: consumer segmentation, new product development, better product placement, new product introduction etc.
For non-experts in statistical analysis, it can be intimidating to determine what method is appropriate for a given application. However, by partnering with a research supplier, such as The Stevenson Company, you can gain a better grasp on which methods of multivariate analysis to use.
What it is
A correlation analysis is used to study the closeness of the relationship between two or more numeric variables. There are several different types of correlations (i.e. positive, negative, simple, partial, multiple, linear or non-linear) as well as several different methods for calculating them. The most common correlation seen in research is linear. For a linear correlation a change in one variable will cause a change in the other variable. A positive correlation exists when one variable increases, while the other one does, too. A negative correlation exists if one variable increases as the other decreases.
The strength of the relationship between data sets is measured using a correlation coefficient. A correlation coefficient is a value between -1 and 1. A value of -1 indicates a perfect negative relationship and 1 a perfect positive relationship. The closer a value is to 1 (or -1) the stronger the relationship between the two data sets. A correlation coefficient of 0 would indicate no relationship.
Applications in market research
Analysts can look for relationships between any two numeric variables. While correlation does not always mean causation, correlation analysis can provide valuable insights for researchers. Whether or not causation is present, correlation does indicate a relationship. That relationship can form the basis for hypothesis and further testing. One frequent application used in research is to find out what matters most to respondents by correlating survey attributes with overall performance or satisfaction. It is also often used for predictions and forecasting.
The chart below shows a positive correlation between two datasets, with a correlation coefficient of 0.986, indicating a strong relationship between the datasets. Positive correlation could exist between many different pairs of variables. Some examples include household income and spending or study time and test scores.
What it is
Researchers use factor analysis as a data reduction technique, consolidating a large number of variables into a handful of independent underlying factors. Using this method, analysts can create groups of variables that make the results more understandable, thus more actionable. By using factor analysis researches can identify trends faster, see patterns developing throughout the dataset, and enable clients to identify commonalities across data points. The assumption is that several variables are related in a way that can be explained by an underlying “factor”. The most common approach is to use correlation matrices (some use covariance) to identify patterns or relationships between variables.
Applications in market research:
In market research factor analysis is frequently used in studies that involve product attributes and perceptions. Assigning attributes to factors results in equal sized groups of attributes. Creating a factor analysis is both an art and a science, grounded in lots of statistics. The resulting factors should be analyzed to determine if the attributes in each factor share a common theme.
If the initial factor solution does not show this internal consistency within factors (e.g. common theme), researchers can investigate alternate factor solutions that may have a different number of factors than the original solution.
Below is a factor solution for 14 attributes, loaded into three factors, showing which specific attributes are part of each factor. The values in this table are called factor loadings and can be interpreted as correlation values. Factor loading values close to -1 or 1 indicate that the attribute strongly influences the factor. Examining the loading patterns will help determine which attributes influence which factors the strongest. Those attributes are then grouped and can be used for further analysis.
What it is
In a cluster analysis, objects are separated into clusters (groups) where the data points of a cluster are similar to each other. The goal is to sort different data points into groups so the data points in the same group are closely associated with each other while the data points belonging to different groups have a low association with each other. This method is typically used in the exploratory phase of research, where no assumptions are present. It helps researchers (and their clients) to identify and define patterns between data elements that were not previously defined. Bringing these patterns to light helps to distinguish and outline structures which might not have been apparent before but give significant meaning of the data once they are discovered, making informed decision-making much easier.
Applications in market research:
Cluster analysis is commonly used to develop market segments that can then allow for better positioning of products and messaging. It can also help a company to better position itself, explore new markets, and develop products that a specific cluster finds relevant and valuable.
The data in the chart below suggests that there are three distinct clusters that likely need to be targeted differently. The Orange segment has high product satisfaction and high brand loyalty, the Blue has high product satisfaction and low brand loyalty, and the third segment (Green) has low product satisfaction and rates in the middle for brand loyalty. These clusters can also be used for further analysis.
Multivariate analyses allow researchers to more fully explore data, which in turn allows them to present their clients with more nuanced findings. While not every question a client asks needs to be answered using multivariate analyses, they can help uncover relationships in the data that might otherwise be overlooked.
Not sure whether you need an in-depth analysis or if a less complex test would suffice? A research partner like The Stevenson Company will be able to guide you through deciding which options will best answer your questions.