SPSSX Discussion

Question 2: Weight question

Classic

List

Threaded

4 messages Options

Neda Faregh

Question 2: Weight question

Hi all,

Here is my question 2 regarding weights ( called SPSS support with this
question - no answers yet.)

I need to apply predetermined survey weights to my data.

SPSS only recognizes full integers as weights and rounds up the weights
(.699, or 1.12) so that most of my data get a weight of 1.

I tried multiplying by 1000 but that inflates the numbers and all tests
become significant.

Any suggestions?

Thank you very much,

Neda

=====================
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

Jerabek Jindrich

Re:Question 2: Weight question

Hi,

> SPSS only recognizes full integers as weights and rounds up the weights
> (.699, or 1.12) so that most of my data get a weight of 1.

No, SPSS does not round nor truncates the weighting variable.

There are two reasons why the wgt variable seems to be rounded to 1:

First - the wgt variable is formated as integer number. Check in the data editor-variable view the format of the wgt variable. But if the thing is only in format that is no problem - SPSS stores the number "precisely", with decimals, even though it is not displayed.

Second - maybe you read the wgt variable from an external file - excell, text ....
Then it could happen that the number is not read correctly, please check the original file and the reading proces.

best
Jindra

> ------------ Původní zpráva ------------
> Od: Neda Faregh <[hidden email]>
> Předmět: Question 2: Weight question
> Datum: 02.7.2008 19:47:36
> ----------------------------------------
> Hi all,
>
> Here is my question 2 regarding weights ( called SPSS support with this
> question - no answers yet.)
>
> I need to apply predetermined survey weights to my data.
>
> SPSS only recognizes full integers as weights and rounds up the weights
> (.699, or 1.12) so that most of my data get a weight of 1.
>
> I tried multiplying by 1000 but that inflates the numbers and all tests
> become significant.
>
> Any suggestions?
>
>
>
> Thank you very much,
>
>
>
> Neda
>
> =====================
> 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
>
>
>

ViAnn Beadle

Re: Question 2: Weight question

Some procedures in SPSS require integer weights and will round or randomly
move up or down to achieve an integer weight when doing calculations within
the procedure.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Jerabek Jindrich
Sent: Wednesday, July 02, 2008 12:55 PM
To: [hidden email]
Subject: Re:Question 2: Weight question

Hi,

> SPSS only recognizes full integers as weights and rounds up the weights
> (.699, or 1.12) so that most of my data get a weight of 1.

No, SPSS does not round nor truncates the weighting variable.

There are two reasons why the wgt variable seems to be rounded to 1:

First - the wgt variable is formated as integer number. Check in the data
editor-variable view the format of the wgt variable. But if the thing is
only in format that is no problem - SPSS stores the number "precisely", with
decimals, even though it is not displayed.

Second - maybe you read the wgt variable from an external file - excell,
text ....
Then it could happen that the number is not read correctly, please check the
original file and the reading proces.

best
Jindra

=====================
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

=====================
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

Hector Maletta

Re: Question 2: Weight question

In reply to this post by Neda Faregh

Neda,
SPSS does not rounds up your weights. In fact, in recent versions you can
choose how fractional weights are to be treated. Values or counts from your
data set are multiplied by the weighting variable, be it integer or not, and
thus the weighted "total" for a single case may be 1.12 or 0.45, and the
weighted sum of 100 cases might be, say, 428.7534876. This is the value
internally stored by SPSS. You can verify this by cutting and pasteing your
output table into an Excel worksheet and allowing for decimals in the Excel
sheet cells, even if the figure of the count or frequency in the output
table appears to be 429 instead of 428.7534876.
There are some problems with fractional weights, though. One of them is that
certain procedures do not accept them (e.g. CATCPA in the Categories
module). Another problem is that different tables may show slightly
different totals due to rounding.
Last but not least: it is true that inflationary weights make all your
results artificially significant. But applying weights that add up to your
original sample size does not necessarily get the true significance level if
the sample design is complex. For those cases you need the COMPLEX SAMPLES
procedure or some other similar software device, to take account of
stratification and clustering in your sampling design. Weighting assumes the
sample is a simple random sample, ignoring the effect of stratification
(which reduces error) and clustering (which increases error). If your sample
is not random, of course, there is no point in applying parametric tests of
significance, based on the properties of random samples.

Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Neda Faregh
Sent: 02 July 2008 14:38
To: [hidden email]
Subject: Question 2: Weight question

Hi all,

Here is my question 2 regarding weights ( called SPSS support with this
question - no answers yet.)

I need to apply predetermined survey weights to my data.

SPSS only recognizes full integers as weights and rounds up the weights
(.699, or 1.12) so that most of my data get a weight of 1.

I tried multiplying by 1000 but that inflates the numbers and all tests
become significant.

Any suggestions?

Thank you very much,

Neda

=====================
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

=====================
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