Quantcast

interpreting t-test results

classic Classic list List threaded Threaded
10 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

interpreting t-test results

Talino Bruno
I have set of results for two different tests taken by the same subjects. I
am using the paired sample t-test in SPSS, I assume that is the right one.
I get a significant difference between the two  test, both are
psychological measures, but I would like to know which test gives me the
higher score i.e which psychological construct is most prevalent in my
sample, the test scores are calculated differently one ranges from 1- 4 and
the other from 24-120. Would I use another analysis to get this answer?

=====================
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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Richard Ristow
At 07:57 PM 4/5/2009, Talino Bruno wrote:

>I have set of results for two different tests taken by the same
>subjects. I am using the paired sample t-test in SPSS, I assume that
>is the right one. I get a significant difference between the
>two  test, both are psychological measures, but I would like to know
>which test gives me the higher score i.e which psychological
>construct is most prevalent in my sample, the test scores are
>calculated differently one ranges from 1- 4 and the other from
>24-120. Would I use another analysis to get this answer?

The paired t-test tells you whether the score consistently increases
or decreases from the first test to the second.

If the scores on the first test are 4 or less, and those on the
second are 24 or more, then the numerical change from the first to
the second is always positive. The t-test will show a clearly
significant result, but that's an artifact; it's meaningless.

You can't test for a significant difference, or even define what it
is, unless the two quantities are measured on the same scale.

=====================
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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

statisticsdoc
In reply to this post by Talino Bruno
Talino,

You would need to compare scales that are based on the same metric in order
to have a meaningful result. The numbers for the second scale may be higher
because the rating scale uses larger numbers, but that does not imply that
levels of some construct are higher.

HTH,

Steve Brand

www.StatisticsDoc.com

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Talino Bruno
Sent: Sunday, April 05, 2009 7:58 PM
To: [hidden email]
Subject: interpreting t-test results


I have set of results for two different tests taken by the same subjects. I
am using the paired sample t-test in SPSS, I assume that is the right one.
I get a significant difference between the two  test, both are
psychological measures, but I would like to know which test gives me the
higher score i.e which psychological construct is most prevalent in my
sample, the test scores are calculated differently one ranges from 1- 4 and
the other from 24-120. Would I use another analysis to get this answer?

=====================
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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Henrik Lolle
So the question is if it's only the coding which is different or if
it's two completely different measures. Can you, in other words, recode
one scale into the other before the t-test?

Best,
Henrik

Quoting Statisticsdoc <[hidden email]>:

> Talino,
>
> You would need to compare scales that are based on the same metric in order
> to have a meaningful result. The numbers for the second scale may be higher
> because the rating scale uses larger numbers, but that does not imply that
> levels of some construct are higher.
>
> HTH,
>
> Steve Brand
>
> www.StatisticsDoc.com
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
> Talino Bruno
> Sent: Sunday, April 05, 2009 7:58 PM
> To: [hidden email]
> Subject: interpreting t-test results
>
>
> I have set of results for two different tests taken by the same subjects. I
> am using the paired sample t-test in SPSS, I assume that is the right one.
> I get a significant difference between the two  test, both are
> psychological measures, but I would like to know which test gives me the
> higher score i.e which psychological construct is most prevalent in my
> sample, the test scores are calculated differently one ranges from 1- 4 and
> the other from 24-120. Would I use another analysis to get this answer?
>
> =====================
> 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
>



************************************************************
Henrik Lolle
Department of Economics, Politics and Public Administration
Aalborg University
Fibigerstraede 1
9200 Aalborg
Phone: (+45) 99 40 81 84
************************************************************

=====================
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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Jims More
In reply to this post by Talino Bruno
Talino,
If you compare ant and elephant, the result is always significant.  Ant is always smaller than elephant.  Your case is analogous to the ant-elephant example.  What do you think?
 
Jims. 

--- On Sun, 5/4/09, Talino Bruno <[hidden email]> wrote:

From: Talino Bruno <[hidden email]>
Subject: interpreting t-test results
To: [hidden email]
Received: Sunday, 5 April, 2009, 11:57 PM

I have set of results for two different tests taken by the same subjects. I
am using the paired sample t-test in SPSS, I assume that is the right one.
I get a significant difference between the two  test, both are
psychological measures, but I would like to know which test gives me the
higher score i.e which psychological construct is most prevalent in my
sample, the test scores are calculated differently one ranges from 1- 4 and
the other from 24-120. Would I use another analysis to get this answer?

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


Enjoy a better web experience. Upgrade to the new Internet Explorer 8 optimised for Yahoo!7. Get it now..
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Martin P. Holt-2
In reply to this post by Talino Bruno
Hi Talino,

Given that one scale measures only from 1-4, is that a measure of the
construct you're interested in, alone ? The other test ranges from
24-120...much bigger...but is it possible to do a similar thing and extract
the one question that measures the construct you're interested in, albeit
from a different angle ? You might then be able to compare the two tests on
that one construct.....although they might still be based on a different
metric (the objection raised so far)....isn't this part of what you are
after ? But it may be too simple to think that you can extract the question
that draws on your construct of interest like this.

Just a thought,
Martin


----- Original Message -----
From: "Talino Bruno" <[hidden email]>
To: <[hidden email]>
Sent: Monday, April 06, 2009 12:57 AM
Subject: interpreting t-test results


>I have set of results for two different tests taken by the same subjects. I
> am using the paired sample t-test in SPSS, I assume that is the right one.
> I get a significant difference between the two  test, both are
> psychological measures, but I would like to know which test gives me the
> higher score i.e which psychological construct is most prevalent in my
> sample, the test scores are calculated differently one ranges from 1- 4
> and
> the other from 24-120. Would I use another analysis to get this answer?
>
> =====================
> 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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Muir Houston-2
In reply to this post by Jims More
is this not what z scores are for?

or am I missing something
 
M


From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Jims More
Sent: 06 April 2009 09:56
To: [hidden email]
Subject: Re: interpreting t-test results

Talino,
If you compare ant and elephant, the result is always significant.  Ant is always smaller than elephant.  Your case is analogous to the ant-elephant example.  What do you think?
 
Jims. 

--- On Sun, 5/4/09, Talino Bruno <[hidden email]> wrote:

From: Talino Bruno <[hidden email]>
Subject: interpreting t-test results
To: [hidden email]
Received: Sunday, 5 April, 2009, 11:57 PM

I have set of results for two different tests taken by the same subjects. I
am using the paired sample t-test in SPSS, I assume that is the right one.
I get a significant difference between the two  test, both are
psychological measures, but I would like to know which test gives me the
higher score i.e which psychological construct is most prevalent in my
sample, the test scores are calculated differently one ranges from 1- 4 and
the other from 24-120. Would I use another analysis to get this answer?

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


Enjoy a better web experience. Upgrade to the new Internet Explorer 8 optimised for Yahoo!7. Get it now..
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Arthur Kramer
In reply to this post by Talino Bruno
After reading some of the other responses, i.e., re-centering the scores so
they have a common mean and standard deviation, and picking an item
indicative of the "construct of interest", I think finding a third, and
maybe a forth, test that you know tests the "constructs of interest" then
correlating the scores from the first two test with the new tests.  That
should give a more precise indication of which construct in more
higher/stronger in the first two.

Arthur Kramer

"...believe half of what you see and none of what you hear."
                  N.Whitfield and B.Strong

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Talino Bruno
Sent: Sunday, April 05, 2009 7:58 PM
To: [hidden email]
Subject: interpreting t-test results

I have set of results for two different tests taken by the same subjects. I
am using the paired sample t-test in SPSS, I assume that is the right one.
I get a significant difference between the two  test, both are
psychological measures, but I would like to know which test gives me the
higher score i.e which psychological construct is most prevalent in my
sample, the test scores are calculated differently one ranges from 1- 4 and
the other from 24-120. Would I use another analysis to get this answer?

=====================
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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Darren Philpott
In reply to this post by Martin P. Holt-2
Hi Talino,

What you are trying to do is compare apples and oranges or elephants and ants to see if they are different.

You can only compare the results of two psychometric tests if they are on the same metric. A major requirement for putting tests on the same metric would be that broadly measure the same construct. If they do, then you can recalibrate the items to produce a common scale and then compare your results. Many programmes are available to do this, but I suggest something like Bilog MG.

Cheers,
Darren


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Martin Holt
Sent: 06 April 2009 10:42
To: [hidden email]
Subject: Re: interpreting t-test results

Hi Talino,

Given that one scale measures only from 1-4, is that a measure of the
construct you're interested in, alone ? The other test ranges from
24-120...much bigger...but is it possible to do a similar thing and extract
the one question that measures the construct you're interested in, albeit
from a different angle ? You might then be able to compare the two tests on
that one construct.....although they might still be based on a different
metric (the objection raised so far)....isn't this part of what you are
after ? But it may be too simple to think that you can extract the question
that draws on your construct of interest like this.

Just a thought,
Martin


----- Original Message -----
From: "Talino Bruno" <[hidden email]>
To: <[hidden email]>
Sent: Monday, April 06, 2009 12:57 AM
Subject: interpreting t-test results


>I have set of results for two different tests taken by the same subjects. I
> am using the paired sample t-test in SPSS, I assume that is the right one.
> I get a significant difference between the two  test, both are
> psychological measures, but I would like to know which test gives me the
> higher score i.e which psychological construct is most prevalent in my
> sample, the test scores are calculated differently one ranges from 1- 4
> and
> the other from 24-120. Would I use another analysis to get this answer?
>
> =====================
> 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

_____________________________________________________________________
This e-mail has been scanned for viruses by Verizon Business Internet Managed Scanning Services - powered by MessageLabs. For further information visit http://www.verizonbusiness.com/uk

No virus found in this incoming message.
Checked by AVG - www.avg.com
Version: 8.0.238 / Virus Database: 270.11.42/2042 - Release Date: 04/05/09 10:54:00

_____________________________________________________________________
This e-mail has been scanned for viruses by Verizon Business Internet Managed Scanning Services - powered by MessageLabs. For further information visit http://www.verizonbusiness.com/uk

=====================
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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: interpreting t-test results

Garry Gelade
In reply to this post by Arthur Kramer
Apologies if I've missed aome of the previous replies, but my 2cents worth -

Let's suppose your two psychological contstructs are extraversion and
neuroticism.  You say you want to find which of the two is "more prevalent
in your sample".  The only ay I can make sense of this is to think in terms
of norm comparisons.  You want to find whether your sample is significantly
different from some reference population on either or both of the
constructs.

A t-test - any paired sample approcah - is no good here as some contributors
have already pointed out.  You need to compare your sample scores with the
scores of some normative reference popuation.  If for example you find the
sample mean for extraversion is not significantly different from the mean of
the reference population norm, but the neuroticism mean does differ  from
that in the reference population, you could conclude that neuroticism was
'more prevalent' in your sample (than in the norm population), but
extraversion was not.

Garry Gelade
Business Analytic Ltd

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Arthur Kramer
Sent: 06 April 2009 14:46
To: [hidden email]
Subject: Re: interpreting t-test results

After reading some of the other responses, i.e., re-centering the scores so
they have a common mean and standard deviation, and picking an item
indicative of the "construct of interest", I think finding a third, and
maybe a forth, test that you know tests the "constructs of interest" then
correlating the scores from the first two test with the new tests.  That
should give a more precise indication of which construct in more
higher/stronger in the first two.

Arthur Kramer

"...believe half of what you see and none of what you hear."
                  N.Whitfield and B.Strong

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Talino Bruno
Sent: Sunday, April 05, 2009 7:58 PM
To: [hidden email]
Subject: interpreting t-test results

I have set of results for two different tests taken by the same subjects. I
am using the paired sample t-test in SPSS, I assume that is the right one.
I get a significant difference between the two  test, both are psychological
measures, but I would like to know which test gives me the higher score i.e
which psychological construct is most prevalent in my sample, the test
scores are calculated differently one ranges from 1- 4 and the other from
24-120. Would I use another analysis to get this answer?

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

__________ NOD32 3989 (20090406) Information __________

This message was checked by NOD32 antivirus system.
http://www.eset.com

=====================
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
Loading...