Nico, Peter, Nassar and Hector, thank you very much for your help! I've tried, and with the RANK command it works.

But I wonder one thing... I don't obtain the exact split on the groups: if I compute the median for each group I have a value; then I create this dummy variable with RANK, but it splits the groups in points different from their medians. It is not too bad for my purposes: the split point is arbitrary. But I don't understand why it is not the median, as supposed.

At 27-03-00, you wrote: >Dear Stefano, >I was too fast with my response. You wanted a dichotomy in each >category; then you can use the following code: >SORT CASES BY category. >SPLIT FILE BY category. >RANK seconds /NTILES(2) INTO speed. > >Now everybody has a code 1 on the variable speed, if (s)he is in the >slowest half of his/her category and 2 otherwise. > >Nico van Ruyven >Fac. of Psychology >Univ. of Amsterdam > > > >Stefano Perlusz wrote: > > > > Dear members, > > > > I've a problem to solve about COMPUTE into a new variabe, and I'm not able > > to solve it by myself. > > > > I have 2 variables: > > - CATEGORY is nominal (1 to 9), it represents the category or type of > > respondent. So, it creates 9 groups in my sample. > > - SECONDS is continuos (>0), represents the time spent to solve a test. > > > > I want to create a dummy variable: > > - SPEED: it should take 0 for slow respondents and 1 for fast respondents. > > Slow respondents are defined as, for each CATEGORY, the first half of > > the respective group (that is, the 50% slower respondents for that group). > > Fast respondents are the others. > > > > What would be the syntax command? > > > > Any help is very much appreciated! > > > > Stefano Perlusz