Alternative ways to stack a dataset

[Jignesh Sutar]

I'm wanting to restructure a dataset such that I need to apply something like this:

VARSTOCASES

  /MAKE Q1 from Q1_01 to Q1_99

  /MAKE Q2 from Q2_01 to Q2_99

  /MAKE Q3 from Q3_01 to Q3_99

  /MAKE Q4 from Q4_01 to Q4_99

  /MAKE Q5 from Q5_01 to Q5_99   X1_01 TO X1_99   Y1_01 TO Y1_99   Z1_01 TO Z1_99.

To achieve this I need to create 3 sets of dummy variables for each Q1 to Q4 (as for each MAKE in VARSTOCASES must contain the same number of variables). To avoid this I'm wondering if the desired result can be reached some other way?

Many thanks in advance,

Jignesh

[David Marso]

You are not describing whatever issue you are having with any clarity.
Try try again!
Beginning? Middle? End?

JSutar wrote

I'm wanting to restructure a dataset such that I need to apply something
like this:

VARSTOCASES
  /MAKE Q1 from Q1_01 to Q1_99
  /MAKE Q2 from Q2_01 to Q2_99
  /MAKE Q3 from Q3_01 to Q3_99
  /MAKE Q4 from Q4_01 to Q4_99
  /MAKE Q5 from Q5_01 to Q5_99   X1_01 TO X1_99   Y1_01 TO Y1_99   Z1_01 TO
Z1_99.


To achieve this I need to create 3 sets of dummy variables for each Q1 to
Q4 (as for each MAKE in VARSTOCASES must contain the same number of
variables). To avoid this I'm wondering if the desired result can be
reached some other way?

Many thanks in advance,
Jignesh

[Andy W]

I can think of no easier way - at least not any easier than building fake variables. One possibility may be to use one big VARSTOCASES - something like

VARSTOCASES /MAKE Q FROM Q1_01 TO Z1_99 /INDEX Id (Q).

And then rebuild the ID to suit your fancy for a CASESTOVARS to reshape the variables back in different columns. Another idea is to make seperate datasets, do the VARSTOCASES in each uniquely and then merge them back together if need be. Both of these will likely take more time and code to accomplish the same end result as the dummy variable approach you mention.

Without any context this ends up being a pretty strange way to store your data though, as for the Q variables you will basically have missing values by "data storage design". I can imagine I'm ignorant - but I've never come across a situation in which that is desirable (and is most of the time an outcome of poor database design/normalization in my experience).

[Rich Ulrich]

Why?

But it looks like you can put out X1_01 to Z1_99 as a
separate file that has only variable Q5 by using one VarsToCases;
put out all the Q variables using the sensible VarsToCases;
and then ADD FILES.

--
Rich Ulrich

---

Date: Tue, 4 Mar 2014 12:25:12 +0000
From: [hidden email]
Subject: Alternative ways to stack a dataset
To: [hidden email]

I'm wanting to restructure a dataset such that I need to apply something like this:

VARSTOCASES

  /MAKE Q1 from Q1_01 to Q1_99

  /MAKE Q2 from Q2_01 to Q2_99

  /MAKE Q3 from Q3_01 to Q3_99

  /MAKE Q4 from Q4_01 to Q4_99

  /MAKE Q5 from Q5_01 to Q5_99   X1_01 TO X1_99   Y1_01 TO Y1_99   Z1_01 TO Z1_99.

To achieve this I need to create 3 sets of dummy variables for each Q1 to Q4 (as for each MAKE in VARSTOCASES must contain the same number of variables). To avoid this I'm wondering if the desired result can be reached some other way?

Many thanks in advance,

Jignesh

[Richard Ristow]

At 07:25 AM 3/4/2014, Jignesh Sutar wrote:

>I'm wanting to restructure a dataset such that I need to apply
>something like this:
>
>VARSTOCASES
>   /MAKE Q1 from Q1_01 to Q1_99
>   /MAKE Q2 from Q2_01 to Q2_99
>   /MAKE Q3 from Q3_01 to Q3_99
>   /MAKE Q4 from Q4_01 to Q4_99
>   /MAKE Q5 from Q5_01 to Q5_99
>                 X1_01 TO X1_99
>                 Y1_01 TO Y1_99
>                 Z1_01 TO Z1_99.

The first question is, what structure do you want the final dataset
to have? Assuming that each of the TO lists represents 100 variables,
as they seem to, it looks like you want to create 400 cases from each
case in the original file; and the new file contains variables Q1,
Q2, Q3, Q4, and Q5.

It's clear enough what the first 100 records from each original
record should be: Q1-Q5 should have the values from what I'll call
the Q1 group (Q1_01 TO Q1_99) through the Q5 group (Q5_01 to Q5_99)
respectively.

But what about the remaining 300 records? You seem to want Q5 to go
on, taking data from the X1 group for 100 records, from the Y1 group
for 100 records, and the Z1 group for the last 100 records. OK; but,
what values should Q1 to Q4 have in those records? Missing? Or,
repeat their values from the initial 100 records, in sequence? Or, what?

The problem isn't the VARSTOCASES syntax; that can be solved, once we
know what effect you're trying to get.

(By the by, it isn't clear that you have identifying information in
your original records. You should, and that identifying information
needs to remain in the restructured data. Also, you should have an
INDEX subcommand on the VARSTOCASES, so you can trace back any record
in the restructured file to which record *and variables* it came from
in the original file.)

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Richard Ristow]

In previous post, for "100" read "99"; for "400" read "396"; etc.

-Sigh
  Richard

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD