cluster analysis speed

I have downloaded the latest trial version of statisXL and am evaluating as an alternative/replacement for systat11.
in systat11 I am able to perform a hierarchical cluster analysis on a 125 by 1764 or so matrix in a reasonable clusetering the 125 rows, in reasonable time. my problem with systat is that the distance measure is supposed to be 1 - the pearson correlation coefficent, so distances should be between 0 and 2. yet i sometimes get distances greater than 2.
I want to compare the output from systat and statistixl using the same distance measure and linkage method
however, i am currently running statistixl on the same data set. i started more than 10 minutes ago. in the meantime i have run the same analysis in systat several times, each one finishing in less than 5 seconds.
i am not really concerned about the speed of statistxl.... but should it be quicker?
task manager tells me it aint responding... has it crashed excel or should i give it more time?

thanks

Comments

  • Hi

    Generally statistiXL performs most tests in only a matter of seconds (or less). The clustering algorithm is quite CPU intensive and a large dataset could, however, result in a prolonged analysis. It is very unlikely that statistiXL has crashed Excel and I would wait a bit longer and see if you get a result.

    Are you trying to cluster 125 items or 1764? My interpretation of your posting is that in systat you are clustering 125 items (i.e. rows) with 1764 measurements for each item (i.e. columns), is this correct? This analysis cannot currently be performed by statistiXL due to Excel's 256 column limit though we are considering allowing the choice between 'data-in-rows' and 'data-in-columns' in a future version to overcome this limitation.

    Cheers

    Alan
  • i am/was trying to cluster 125 columns each with 1764 measurements for each column, ie 1764 rows....
    which statistiXL cant handle right now...
    bummer!
    i would STRONGLY recommend that y'all allow the option of 'data-in-rows' and 'data-in-columns' 'cos some people, like me, are gonna run into excels column limit.

    thanks for quick your response Alan

Sign In or Register to comment.