I am conducting a study into the US airline industry for the year 2004 up to 2007. I started my analysis by using the hierarchical and K-means cluster analysis to group each airline carrier from my sample of 36 carriers to one of the four clusters. I did so for every year, which provide very consistent results over the years, i.e. only a few individual cases were allocated to a different cluster comparing all four years.
At this point I'd like to create a data sheet in which all the 36 carriers are included for each time period. Therefore the total number of observations will be 36 x 4 = 144 cases for this new data sheet. In order to control for these different time periods I included K-1 = 4-1 dummy variables in which each year will be compared to the final year 2007. So the dummy variable for 2004, the first year, has a 1 for all 36 cases from 2004 and a 0 for all 108 remaining cases of all other years. In addition a dummy variable for 2005 and 2006 are included into the data sheet.
As explained above, I made a classification of clusters for the 36 cases in my sample for each year. Now I'd like to create dummy variables that explains cluster membership. Although the structure of these clusters are not identical for all years. E.g. in 2004 I have 4 clusters with 7,7,12,10 cases in each cluster, while in 2006 I have 4 clusters with 7,6,10,13 cases in each cluster. Therefore I run into the difficulty how to create the dummy variables for cluster membership.
E.g. should I start with ONE nominal variable, for all years, with values 1,2,3,4 that stands for each cluster, and from that create 3 dummy variables? OR, should I start with ONE nominal variable for EACH year separately with values 1,2,3,4 and from that create dummy variables for cluster membership per year?
I am confused with the latter option because I already made dummy variables representing each year. Could somebody help me with this question? If additional information, or the data sheet is required to provide an adequate answer please ask.
Thanks in advance for your support.