One way to select possible number of clusters in hierarchical methods is to identify the relatively large changes in distance as number of clusters change (see table below). When I look at the actual distance measure from the table (clustering history below dendogram), there is a bigger jump from 3 clusters to 4 than from 4 to 5 (smaller), suggesting 4 may be the right decision. However, when you look at the dendogram the jump from 4 to 5 looks bigger. I’m not sure why that is happening. Can anyone please clarify this?
