Perform sensitivity analysis for cluster command results .-
^clustsens^ varlist [^if^ exp] [^in^ range] [, reps(100) k(3) seed(string) d > etail type("kmeans") distopt("L2") dots ]
Description ----------- clustsens is a routine to do a sensitivity analysis of the @cluster@ command.
The cluster command groups cases based on the values of a variable, or the mean/median of a group of variables. The results vary depending on the case the process starts at, which is set by the random number seed. The amount of variation is related to the number of variables being averaged together. The more variables averaged, the less variation (shades of the central limit theorem). At any rate, there will, very likely, be variation - sometimes large variation - in the results. That's what this routine investigates. It simply runs the cluster command multiple times and records the result. The seed is saved (and displayed if the detail option is specified) so that a specific result can be replicated.
Options: -------- ^reps(^#^)^ specifies the number of times the cluster command is repeated. The default is 100.
^k(^#^)^ specifies the number of groups (see help for the @cluster@ command). The default is 3.
^seed(^"string"^)^ specifies the random number seed to use at the start. The default is "123456789".
^type(^"string"^)^ indicates the type of averaging (see help for the @cluster@ > command). The default is "kmeans".
^distopt(^"string"^)^ Specifies how the distance between the cases is calculate > d (see help for the cluster command). The default is "L2".
^detail^ specifies that the detail for each replication will be shown.
^dots^ will print a dot for each replication (shows the command is working...)
Other routines called --------------------- @matsort@ must be installed for clustsens to function
Examples -------- . ^clustsens^ var1 var2 , detail
Author: Paul Millar www.ucalgary.ca/~pemillar/stata.htm pemillar@@ucalgary.ca See also: --------- Online: help for @cluster@, @set@ (if installed)