Working with large data sets In normal usage Survo is fairly verbose by telling on the screen various details about current events when commands and operations are activated. By default it also saves results of many operations as text in a special output file (text file) given by 'eout' in SURVO.APU (default is <Survo>\TMP\RESULTS). When working with large data sets (especially in simulation experiments) these subsidiary effects may be very time-consuming. Then, to spare time, certain special settings are recommended. Saving results to a text file is disabled by the command OUTPUT - (See OUTPUT?). Unessential display of intermediate results in many operations is removed by using a specification PRIND=0 and/or setting prind=0 in SURVO.APU. In certain operations RESULTS=0 also eliminates unnecessary output. See an example on the next page! ....................................................................... Example: Correlations with different output options Creating a sample of 10000 observations from a 200-dimensional multivariate normal distribution: p=200 r=0.7 MAT R=IDN(p,p,1-r)+CON(p,p,r) MNSIMUL R,*,TEST,10000,0 ....................................................................... TIME COUNT START / Calculations on a 366 MHz Pentium CORR TEST / PRIND=1 / In all cases results saved in CORR.M and MSN.M TIME COUNT END 39.870 / Temporary display of correlations took 20 sec. ....................................................................... OUTPUT - / Activate this before CORR! TIME COUNT START CORR TEST / PRIND=1 TIME COUNT END 10.270 / Listing of case numbers took almost 4 sec. ....................................................................... OUTPUT - / Activate this before CORR! TIME COUNT START CORR TEST / PRIND=0 TIME COUNT END 6.370 / Time when unessential displays are removed