3
$\begingroup$

If I have highly dimensional data and perform k means clustering, I can end up with clusters that differ more on some axes than others.

SPSS provides the option to return an ANOVA for each variable, which shows if the variance is significant between clusters. Is it possible to break this down into pairs of clusters?

For instance, I might have 8 variables. If I end up with 3 clusters, I can see that there is a significance difference between them with regards to several variables. What I can't see is whether this effect is driven primarily by the difference between cluster A and B, B and C or A and C.

Is there a way to do this in SPSS? If not, what tools will allow me to? (I have dabbled in R previously.)

$\endgroup$

1 Answer 1

2
$\begingroup$

This is of course possible in SPSS in several procedures. As the most handy tool to visually explore importances of variables in separating this or that cluster I recommend you AIM command (available only through syntax). AIM appeared around SPSS version 13. An example:

cluster       v1       v2 

       1    .2231   -.5798 
       1   -.1558   -.3753 
       1   -.8417   -.4448 
       1   -.1272   -.2229 
       1   -.8788   -.3645 
       1  -1.1993   -.4423 
       1   -.1588   -.1964 
       1   -.5470   -.2505 
       1   -.0336   -.5034 
       1   -.1905   -.3832 
       1   -.3292    .0074 
       1   -.3182   -.2921 
       1   -.0638    .0615 
       1   -.2502   -.7400 
       1   -.0685   -.0429 
       1   -.1812   -.2995 
       1   -.3448   -.8615 
       1   -.3165   -.8266 
       1   -.4237   -.4987 
       1   -.0696    .0052 
       1   -.7209   -.2079 
       1   -.1317   -.4034 
       1  -1.0060   -.2763 
       1   -.8393   -.2328 
       1   -.4828   -.3131 
       1   -.1349   -.3184 
       1   -.5628   -.7562 
       1   -.0377   -.4653 
       1   -.3378   -.2142 
       1   -.3645   -.5174 
       1   -.3462   -.5563 
       1   -.4714   -.4909 
       1   -.5881   -.3015 
       1   -.5467   -.2004 
       2    .1798    .8334 
       2    .3564    .4477 
       2    .2853   1.0951 
       2    .2771    .9481 
       2    .0873    .2397 
       2   -.0341    .3141 
       2    .6199    .3630 
       2   -.1225    .2946 
       2   -.1339    .5866 
       2    .1811    .4852 
       2    .3209    .0284 
       2    .3955    .4883 
       2    .1535    .5964 
       2    .5768    .7119 
       2    .4287    .8511 
       2    .0981    .7097 
       2    .2050    .1230 
       2   -.0332    .5411 
       2    .3924   1.3310 
       2    .0739    .4214 
       2    .2895    .5793 
       2    .4332   1.1919 
       2    .3213    .7937 
       2    .4064    .6395 
       2    .2301    .7870 
       2    .0613    .5863 
       2    .2007    .3988 
       2    .2399    .5613 
       2    .1089    .1089 
       2   -.1748    .5251 
       2    .2514    .1649 
       2    .1680    .1657 
       2    .3827    .5861 
       2   -.0621    .4763 
       2    .2319    .7849 
       2    .3100    .5979 
       2    .1499    .2342 
       2    .0216    .1014 
       2    .5882    .2807 
       2    .7664    .7398 
       2    .1615    .3927 
       2    .1519    .1471 
       2    .2836    .4367 
       2    .1595    .6316 
       3   -.1826  -2.0091 
       3   -.5102  -1.5094 
       3    .5034  -1.7920 
       3    .4083  -1.8324 
       3    .0334  -1.4827 
       3    .5465  -1.9371 
       3   -.5001  -1.2950 
       3    .2499  -1.8115 
       3    .1666  -1.9963 
       3    .0307  -2.0253 
       3   -.0097  -1.2936 
       3    .2918  -1.6235 
       3   -.1456  -1.4711 
       3   -.3286  -1.4145 
       3    .0706  -1.3348 
       3   -.0333  -1.0373 
       3    .1842  -1.7344 
       3    .2296  -1.0626 
       3   -.0070  -1.7084 
       3    .0243  -2.0720 
       3    .0838  -1.9156 
       3   -.1441  -1.5803 
       3    .0586  -1.2546 
       3   -.2590   -.9835 
       3   -.5447  -1.1362 
       3   -.5730  -1.2252 
       3   -.0953  -1.4002 
       3   -.1344  -1.2260 
       4   1.7083   -.8018 
       4   1.4596  -1.5330 
       4   1.3272  -1.3483 
       4   1.5935  -1.2130 
       4   1.5339  -1.4812 
       4   1.4183  -1.3126 
       4   1.4058  -1.1323 
       4   1.5627  -1.2020 
       4   1.8812  -1.0260 
       4   1.5191  -1.5860 
       4   1.6925  -1.2371 
       4   1.6398  -1.4556 
       4   1.4500  -1.0027 
       4   1.5519   -.9984 
       4   1.3867  -1.5675 
       4   1.9775  -1.1251 
       4   1.6405   -.7982 
       4   1.4092  -1.0957 
       4   1.6009  -1.1614 
       4   2.3062   -.8913 
       4   1.5712   -.9117 
       4   2.0490  -1.2672 
       4   1.4734   -.7397 
       4   1.5605  -1.2318 
       4   1.9052  -1.1109 
       4   1.1757  -1.3380 
       4   2.0886   -.7270 
       4   1.4767   -.8315 
       4   1.1704  -1.5511 
       4   1.3156  -1.3176 
       4   1.2021  -1.5661 
       4   1.6448  -1.5968 
       4   1.5025  -1.4033 
       4   1.6077  -1.0214 
       4   1.4057  -1.1644 
       4   2.1808   -.4811 
       4   1.6106  -1.0972 
       4   1.4535  -1.5626 
       4   1.9538  -1.1485 
       4   1.5477  -1.8905 
       4   2.1102  -1.1612 
       4   1.5932  -1.7512 
       4   1.4653  -1.3547 
       4   1.3127  -1.1288 
       4   1.5494  -1.0847 
       5    .6644   -.8506 
       5    .5331   -.5829 
       5    .8806  -1.1934 
       5    .9727   -.9446 
       5    .5241   -.6445 
       5   1.0435  -1.0571 
       5    .8547   -.7993 
       5    .9354   -.6720 
       5    .7424   -.7735 
       5    .8647   -.9777 
       5   1.0462   -.7624 
       5    .6158   -.6192 
       5    .6948  -1.2458 
       5    .9593   -.7973 
       5    .8625   -.9526 
       5    .8100  -1.1779 
       5    .5450   -.9904 
       5    .4747   -.8331 
       5    .8437   -.8683 
       5    .6862  -1.1216 
       5   1.1181  -1.0946 
       5    .8396  -1.3752 
       5    .9554   -.8151 
       5    .7777   -.7837 
       5    .6653   -.5683 
       5   1.0094  -1.0178 
       5   1.1374  -1.0900 
       5   1.1849   -.6677 
       5    .3905   -.6421 
       5   1.1475   -.4460 
       5   1.0077  -1.3488 
       5    .5587   -.2977 

*Display clusters together. Compares degree of apartness of clusters on a specific variable.
AIM cluster
 /CONTINUOUS v1 v2
 /PLOT IMPORTANCE (X=GROUP Y=TEST).

*Display variables together. Compares degree of importance of variables for a specific cluster.    
AIM cluster
 /CONTINUOUS v1 v2
 /PLOT IMPORTANCE (X=VARIABLE Y=TEST).

Read Command Syntax Reference document (available in Help and as pdf file) for more about this command.

$\endgroup$
2
  • $\begingroup$ Pardon my ignorance, but could you give me an example to get me started? $\endgroup$
    – Tom Wright
    Commented May 3, 2013 at 11:14
  • 1
    $\begingroup$ @Tom, example given $\endgroup$
    – ttnphns
    Commented May 3, 2013 at 12:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.