Hi,
I just discovered which GREAT progress this software made over the
years. Thanks very much for that to the contributors!
I want to clarify this before sending bug reports: In the archive of
this list I found the thread with the same subject from July 2014. I
think that the implementation of MEANS TABLES is buggy. Particularly the
meaning of "MEANS TABLES = a by b by c" seems mixed up with "MEANS
TABLES = a by b c".
Given this example, which is slightly adjusted from the manual one's
(the manual states that each of the three subcommands produces only one
table):
get file = 'C:\Programme\PSPP\share\pspp\examples\hotel.sav' .
MEANS TABLES =
v1 v2 v3 BY v4 /
v1 v2 BY v4 v5 /
v1 BY v4 BY v5 .
The first subcommand produces two tables, both grouped in three blocks
v1 v2 v3. The first one has results for single values of v1, v2, v3, the
second one has totals. These are the same numbers as in SPSS, but SPSS
arranges them in one table and has v1, v2, v3 as columns and mean,
count, stddev as rows.
The second subcommand also produces two tables, also both grouped in two
blocks v1 v2, also one for details and one for totals (no subtotals).
The details are for single values of v4 and v5 combined, I think this is
what the manual describes "a categorical variable formed by the
combination of y and z", but which references the _third_ subcommand.
In SPSS the _third_ subcommand produces one table with the same detail
numbers, again with total and subtotals.
The third subcommand produces three tables, ungrouped, two for details
and one for total. In SPSS the _second_ subcommand produces two tables
with the same numbers.
Using psppire.exe 0.8.4-g012d99, 64 bit from sourceforge in Windows 8.
Using SPSS for Windows Release 8.0.0 (11 May 1998) in Windows
2000/Oracle VM Virtualbox. :-) :-(
--
Kai Borgolte, Bonn
_______________________________________________
Pspp-users mailing list
Pspp-users@gnu.org
https://lists.gnu.org/mailman/listinfo/pspp-users
