Weighting Question - Ratio Data

[Colin Valdiserri]

Hello everyone,



I have a data set that I need to weight using three variables. Each of the
three variables is not categorical, but ratio data. The respondents were
asked to enter a percentage for each of the three variables and the sum of
them yields 100%.



I have weighted data before, but using categorical data such as demographics
and am not quite sure how to go about it with the ratio data.



Typically I would use the following syntax or go to rim weighting.



If (Q2_1=1) wgt=29.1/50.42.

If (Q2_2=2) wgt=26.7/31.68.

If (Q2_3=3) wgt=44.2/17.90.



But since the data is not categorical/nominal, I am unsure how to approach
this syntax. All help is appreciated



Thanks,



Colin

=====================
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]

Colin,
What is the weighting variable supposed to mean or achieve?
Suppose a respondent answered with three percentages, say 40, 40 and 20,
while another respondent responded 20, 30 and 50. What you intend to do with
these numbers? How would they influence the "weight" of each respondent in
the final count of responses?
One possibility is that you already know the "true" percentages, and want to
give each respondent more weight if they are nearer the true percentages, or
some other such idea. But I cannot imagine a situation in which such would
be meaningful.
Weighting by an interval variable, on the other hand, is quite usual. For
instance, very often you have a variable representing the inverse of the
sampling ratio, i.e. how many cases in the population are represented by
each case in the sample. If the first case came from a district were one
case out of every 500 were selected, its weight would be 500, while cases
from some other district would have a different weight according to the
sampling ratio used in each district.
If you need to weight by two or more interval variables, this is also
doable, through so-called "rim" weighting or some similar procedure, though
SPSS does not do it by itself.
Lastly, notice that one3 of the three percentages is not necessary, as it is
determined by the other two: you need only use two. But you better explain
yourself in more detail.

Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Colin Valdiserri
Sent: 04 June 2008 13:50
To: [hidden email]
Subject: Weighting Question - Ratio Data

Hello everyone,



I have a data set that I need to weight using three variables. Each of the
three variables is not categorical, but ratio data. The respondents were
asked to enter a percentage for each of the three variables and the sum of
them yields 100%.



I have weighted data before, but using categorical data such as demographics
and am not quite sure how to go about it with the ratio data.



Typically I would use the following syntax or go to rim weighting.



If (Q2_1=1) wgt=29.1/50.42.

If (Q2_2=2) wgt=26.7/31.68.

If (Q2_3=3) wgt=44.2/17.90.



But since the data is not categorical/nominal, I am unsure how to approach
this syntax. All help is appreciated



Thanks,



Colin

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

[Peck, Jon]

A few comments below

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Hector Maletta
Sent: Wednesday, June 04, 2008 9:18 PM
To: [hidden email]
Subject: Re: [SPSSX-L] Weighting Question - Ratio Data

Colin,
What is the weighting variable supposed to mean or achieve?
[>>>Peck, Jon] [snip]

Weighting by an interval variable, on the other hand, is quite usual. For
instance, very often you have a variable representing the inverse of the
sampling ratio, i.e. how many cases in the population are represented by
each case in the sample. If the first case came from a district were one
case out of every 500 were selected, its weight would be 500, while cases
from some other district would have a different weight according to the
sampling ratio used in each district.

[>>>Peck, Jon]
[>>>Peck, Jon] Note that proper treatment of probability-of-selection weights is often different from replication weights, at least as far as standard errors etc.  The SPSS Complex Samples option is used to address complex sampling schemes.

If you need to weight by two or more interval variables, this is also
doable, through so-called "rim" weighting or some similar procedure, though
SPSS does not do it by itself.

[>>>Peck, Jon]
[>>>Peck, Jon] SPSS does have a module on SPSS Developer Central (www.spss.com/devcentral) that does rim weighting.  It requires that programmability be installed.  It is possible to rig this up with GENLOG, but that is pretty complicated, so the rake module on Developer Central is a much easier way to go.

-Jon Peck


Lastly, notice that one3 of the three percentages is not necessary, as it is
determined by the other two: you need only use two. But you better explain
yourself in more detail.

Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Colin Valdiserri
Sent: 04 June 2008 13:50
To: [hidden email]
Subject: Weighting Question - Ratio Data

Hello everyone,



I have a data set that I need to weight using three variables. Each of the
three variables is not categorical, but ratio data. The respondents were
asked to enter a percentage for each of the three variables and the sum of
them yields 100%.



I have weighted data before, but using categorical data such as demographics
and am not quite sure how to go about it with the ratio data.



Typically I would use the following syntax or go to rim weighting.



If (Q2_1=1) wgt=29.1/50.42.

If (Q2_2=2) wgt=26.7/31.68.

If (Q2_3=3) wgt=44.2/17.90.



But since the data is not categorical/nominal, I am unsure how to approach
this syntax. All help is appreciated



Thanks,



Colin

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

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