The simplest method of data input is from a text file, with the values separated by spaces or tab characters. The data should be a square matrix, with the rows corresponding to the cases and the columns corresponding to the variables. The cluster analysis will find groups of cases which are similar - if you are interested in clustering your variables, click here. Another way of reading data is from an Excel spreadsheet. We recently improved our Excel interface so that, for example, a spreadsheet of 5000 rows and 10 columns is read in 2 seconds. Your data can contain missing values, which should be coded with a unique missing value code. You may have different types of variables, in which case it will be necessary to specify their types. You can do this in ClustanGraphics after you have read your data. Another decision is whether or not to transform your variables prior to clustering. |