Hi, I ran a card-sorting experiment (free sorting) where I asked subjects to
group a set of 100 items into groups of similarity.  The subjects decided
the number of groups they wanted to create.  I am going to run cluster
analysis to see the number of clusters that emerged from the data.  I will
then develop items to measure the groups that emerge.  So essentially, the
emerged groups are the constructs for my questionnaire.

I was asked how do I know I have enough sample size (i.e., participants) to
feel comfortable with the result of the cluster analysis.  I have checked
several books on cluster analysis, none discusses the sample size issue.
So, I would appreciate it if anybody can give me some pointers.  Is there
any rule of thumb governing the minimum number of sample size (i.e.,
subjects) required to run a cluster analysis?  Any reference out there
discussing this issue?

Thanks.

Reply via email to