- When the actual data points are given ( the previous example),
it is recommended that we use FA rather that MDS because
- MDS solution may not be as robust as the FA solution
- FA solution has a more rigorous mathematical framework
- When only the dissimilarity matrix is given, we should use MDS
and not use FA becuase
- These distances do not help in conveying relationships
regarding the covariance and/or correlation matrix
- These do not help us estimate the number of factors
- Before we perform FA or MDS, we need sufficient reason to
believe that there is some form of clustering among the data