paired vs independent samples t-test

[Humphrey Paulie]

Dear all,
in pre-test post-test studies where only one group in involved for comparing the means before and after treatment "paired samples t-test" is required. is that right?
i was just wondering how can independent samples t-test in such a situation change the results. I mean is it possible to find a mean difference significant with a pairde t-test and insinificant with an independent t-test?
cheers
Humphrey




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

[Oliver, Richard]

I think the answer is yes, it is possible for the same mean difference to be significant for a paired-samples test but not significant for an independent samples test.

set seed 123456789.
input program.
loop #i=1 to 100.
*post is close to pre, but always greater.
compute pre=rv.normal(50,10).
compute post=pre+rv.normal(.5,.1).
end case.
end loop.
end file.
end input program.
T-TEST PAIRS=pre WITH post (PAIRED).
*now restructure for independent samples test.
VARSTOCASES
  /ID=id
  /MAKE score FROM pre post
  /INDEX=group(2).
value labels group 1 'pre' 2 'post'.
T-TEST GROUPS=group(1 2)
  /VARIABLES=score.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Humphrey Paulie
Sent: Tuesday, September 09, 2008 2:34 PM
To: [hidden email]
Subject: paired vs independent samples t-test

Dear all,
in pre-test post-test studies where only one group in involved for comparing the means before and after treatment "paired samples t-test" is required. is that right?
i was just wondering how can independent samples t-test in such a situation change the results. I mean is it possible to find a mean difference significant with a pairde t-test and insinificant with an independent t-test?
cheers
Humphrey




=======
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

[Burleson,Joseph A.]

No, the paired t-test will always be as or more significant. If the correlation between the pre- and the post-test is exactly r = 0.000, then the two tests will be equal. If the r = 1.000, then the difference scores between the pre- and the post-test will all be identical, meaning that the SD of the difference scores will be zero. In this case, all differences will be infinitely significant.

Joe Burleson

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Humphrey Paulie
Sent: Tuesday, September 09, 2008 3:34 PM
To: [hidden email]
Subject: paired vs independent samples t-test

Dear all,
in pre-test post-test studies where only one group in involved for comparing the means before and after treatment "paired samples t-test" is required. is that right?
i was just wondering how can independent samples t-test in such a situation change the results. I mean is it possible to find a mean difference significant with a pairde t-test and insinificant with an independent t-test?
cheers
Humphrey




=======
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