Business Research - Multi-Dimensional Scaling
Multi-Dimensional Scaling
|
Multi-Dimensional
Scaling Multi-dimensional scaling (MDS) is a statistical technique
that allows researchers to find and explore underlying themes, or dimensions,
in order to explain similarities or dissimilarities (i.e. distances) between
investigated datasets. You can analyse any kind of
similarity or dissimilarity matrix using multi-dimensional scaling. Plotting
these data sets on a multi-dimensional scale allows for easier interpretation
and comparison by researchers. A possible
example of when multi-dimensional scaling (MDS) might be used is if we have
six utility companies and we want to understand how they are considered
differently by respondents. We would invite consumers to complete a survey in
which each of the six companies would be paired with each of the others, and
the respondents would be asked in a series of questions how similar they
believe them to be, for a number of attributes. Examples of attributes may
be: quality, service and price. We can specify that we would like to reproduce the
data on two dimensions. As a result of the MDS analysis we would get an
output for each attribute rated that shows the two-dimensional representation
of how similar or different the companies are viewed. This makes the data
much easier to look at and gives the observer a clearer sense of how
different each company is. This can be used in brand positioning, identifying
if work is needed to make a particular utility company’s brand more unique in
the specific market place. In market research, multi-dimensional scaling is
often used to plot data such as the perception of products or brands; this
will display both the number and nature of the dataset in an easy to
interpret, visual way. This being said, any kind of data with meaningful
similarities or distances can be displayed using multi-dimensional scaling.
Classical multi-dimensional scaling portrays similarities and dissimilarities
between pairs of items on a coordinate matrix. Interpreting the dimensions is made simple by
plotting them in a two-dimensional coordinate matrix (scatter graph);
three-dimensional data can also be displayed graphically but is more complex
to read. Once plotted, researchers can better explore the data and any
patterns or clusters occurring. Plotting more than two or three dimensions is
possible, but it is understood that with each additional dimension, the
output matrix becomes more difficult to interpret. |