|
|
Hi,
I observe some uncertainty (checked on SPSS 15, 20) from SPSS' side
as whether a declared dataset which is empty - because saving out
data to it was unsuccesful - should be eligible to activate. This is
an important issue. Compare 3 following snippets.
Let's have working dataset x with variables v1 v2 v3 v4 v5.
A)
DATASET ACTIVATE x.
DATASET DECLARE y.
REGRESSION
� /DEPENDENT w1
� /METHOD= ENTER w2 w3 w4 w5 /*Error: wrong variable names*/
� /OUTFILE= COVB(y).
DATASET ACTIVATE y.
VARIABLE LEVEL all (ORDINAL).
� Still, DATASET ACTIVATE y does bring forward the empty dataset y.
(And so, VARIABLE LEVEL command does not spoil dataset x.)
B)
DATASET ACTIVATE x.
DATASET DECLARE y.
MATRIX.
DO IF 1=2. /*False*/
SAVE 99 /OUTFILE= y. /*so this won't be done*/
END IF.
END MATRIX.
DATASET ACTIVATE y.
VAR LEVEL all (ORDINAL).
In this case, saving doesn't take place because of condition. Still,
even in this case DATASET ACTIVATE y does bring forward the empty
dataset y.
(And so, VARIABLE LEVEL command does not spoil dataset x.)
C)
DATASET ACTIVATE x.
DATASET DECLARE y.
CORRELATE w1 w2 w3 w4 w5 /*Error: wrong variable names*/
� /MATRIX= OUT(y).
DATASET ACTIVATE y.
VAR LEVEL all (ORDINAL).
This case is like (A): saving to y doesn't take place because of the
error. However, unlike (A) or (B), DATASET ACTIVATE y doesn't work,
issuing message: "Dataset y has no data. Execution of this command
stops". As the consequence, VARIABLE LEVEL command - which was
intended for dataset y - acts on dataset x instead, spoiling it.
I believe that behaviour in cases (A) and (B) is correct, and in
case (C) it is malignant. If so, a declared but still empty dataset
should be eligible to activate, and thus,
DATASET DECLARE y.
DATASET ACTIVATE y.
should work and should NOT produce message "Dataset y has no data.
Execution of this command stops".
|
Locked
