Date:         Wed, 31 Oct 2007 14:29:50 -0600
Reply-To:     ViAnn Beadle <vab88011@gmail.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         ViAnn Beadle <vab88011@gmail.com>
Subject:      Re: T-test pair and Graph
Comments: To: behnaz shirazi <mehr59@hotmail.com>
In-Reply-To:  <BAY123-F26771D65AD59220743130CD1930@phx.gbl>
Content-Type: text/plain; charset="us-ascii"

A mean is not a paired t-test.

Many procedures in SPSS allow you to specify that you want missing value
exclusion to be listwise, pairwise (for things that compare or relate pairs
of variables), or on a variable by variable basis. GGRAPH provides listwise
and variable by variable exclusion because it has no real concept of pairs
of variables.

Usually if when plotting multiple variables, the user is doing so because
the user wants to compare all of them--in this instance, listwise deletion
is the best choice. You, however, want to put all of your variables into a
single graph but treat them as pairs--IMHO this is a mistake. If you want
the mean of all six variables in a single chart, then you think that they
must be related at some higher level--listwise deletion is the appropriate
thing to do.

If you want pairs of non-missing cases within a single chart, what you
should do is make sure that if either value in the pair is missing, it is
missing for both and then choose individual treatment of missing values for
all 6 variables.

Something like this untested logic should work:
* start by copying to new variables for safety's sake.
Do repeat oldvars = v1 to v6/newvars=newv1 to newv6.
Compute newvars=oldvars.
End repeat.

Do if missing(newv1) or missing(newv2).
Recode newv1 newv2 (ALL=SYSMIS).
End if.

... do the same for newv3,newv4 and newv5,newv6.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
behnaz shirazi
Sent: Wednesday, October 31, 2007 1:32 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Re: T-test pair and Graph

Thank you very much for your useful information just the question I have
still in my head is about T-test and mean.
If I consider following situation:(Just for one pair)


  V1                              V2

10                                 .
30                                20
20                                30
10                                10
..                                   20

Here I have 2 missing and the number of cases in T-test is 3 for both ,while
when I use mean I have 4 case for V1 and 4 case for V2 as it is seeing them
separately.

The other pairs can have different number of cases as well.The user didn't
enter the value and there are some missing value which I don't need them.

To me the correct number is 3 . That's why when I come to graph I face
problem!

At the moment as a solution I am getting new variables and copy the value of
V1 and V2 if they are not missing as follow(I get rid of missing ones):

V11                        V22
30                          20
20                          30
10                          10

Then I put these new variables in graph.



>From: ViAnn Beadle <vab88011@gmail.com>
>Reply-To: ViAnn Beadle <vab88011@gmail.com>
>To: SPSSX-L@LISTSERV.UGA.EDU
>Subject: Re: T-test pair and Graph
>Date: Wed, 31 Oct 2007 11:51:16 -0600
>
>That is NOT a paired t-test. A paired t-test compares the differences
>between observed values on a case by case basis.
>
>If you want to just place multiple variables on a chart, open Chart Builder
>and drag the 6 variables to the Y-axis and drop them on the + box. If the
>variables are scale this will automatically specify the mean.
>
>If you want some additional separation between each set it gets quite a bit
>more complicated and its probably easier to use the CASETOVARS command to
>create three sets of cases for each pair of variables:
>
>VARSTOCASES
>   /ID=id
>   /MAKE trans1 FROM v1 v3 v5
>   /MAKE trans2 FROM v2 v4 v6
>   /INDEX=Index1(trans1)
>   /KEEP=
>   /NULL=KEEP.
>
>
>You'll probably have to fix up the variable labels for your new pair of
>variables (trans1 and trans2) and provide the appropriate value labels for
>the index variable which will have values 1, 2, and 3.
>
>Once the file is restructured, within the Chart Builder drag your trans1
>and
>trans2 to y, go to the groups tab and choose column panels, then add your
>index variable to that. That will give you separation between the three
>sets
>and a set of labels for each pair.
>
>-----Original Message-----
>From: behnaz shirazi [mailto:mehr59@hotmail.com]
>Sent: Wednesday, October 31, 2007 10:33 AM
>To: vab88011@gmail.com; SPSSX-L@LISTSERV.UGA.EDU
>Subject: Re: T-test pair and Graph
>
>I have 6 variables which I want to compare their mean as a pair.I need bar
>graph in a way I can have each pair next to the other.
>Let's say:
>
>Var1 Var2    Var3 Var4   Var5 Var6
>
>I don't know how I can implement T-test in graph .For fixing this problem I
>got another new 6 variable because I wanted to get rid of missing ones and
>now I am applying them in GPL.But maybe
>there is a more professional way.
>
>Thanks
>
> >From: ViAnn Beadle <vab88011@gmail.com>
> >Reply-To: ViAnn Beadle <vab88011@gmail.com>
> >To: SPSSX-L@LISTSERV.UGA.EDU
> >Subject: Re: T-test pair and Graph
> >Date: Tue, 30 Oct 2007 15:47:51 -0600
> >
> >Describe the graph you want.
> >
> >-----Original Message-----
> >From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
> >behnaz shirazi
> >Sent: Tuesday, October 30, 2007 12:38 PM
> >To: SPSSX-L@LISTSERV.UGA.EDU
> >Subject: T-test pair and Graph
> >
> >Hello Dear All,
> >
> >I wonder how I can implement T-test pair in GPL.I want to show the result
> >as
> >a graph.
> >
> >Any suggestion would be appreciated.
> >
> >Thank you
> >Behnaz
> >
> >_________________________________________________________________
> >Express yourself instantly with MSN Messenger! Download today it's FREE!
> >http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
> >
> >=====================
> >To manage your subscription to SPSSX-L, send a message to
> >LISTSERV@LISTSERV.UGA.EDU (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
> >LISTSERV@LISTSERV.UGA.EDU (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
>
>_________________________________________________________________
>Don't just search. Find. Check out the new MSN Search!
>http://search.msn.click-url.com/go/onm00200636ave/direct/01/
>
>=====================
>To manage your subscription to SPSSX-L, send a message to
>LISTSERV@LISTSERV.UGA.EDU (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

_________________________________________________________________
Don't just search. Find. Check out the new MSN Search!
http://search.msn.click-url.com/go/onm00200636ave/direct/01/

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (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
LISTSERV@LISTSERV.UGA.EDU (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