|
|
Dear Rich Ulrich;
thanks for your reply. I only mentioned 3 groups cause i thought WOULD BE enough for the sake of illustration. Indeed, I have 4 groups, A, B, C and D. The problem is that A and B are controlling groups (represent native speakers of MOTHER Language,L1, and TARGET Language, L2), B and C are learner groups. So, in reporting data, if I say that the controling groups behave in the same way (A-B INSIGNIFICANT) and C behaves like A (influenc'ed by L1), but not B (not influenced by L2) when A and B are equal!!! It would seem a contraduction.
In my study, I think, ,I might be mistaken, that only three possible outcomes can be possible.
A=B; B/C=A & B; I conclude that there's positive influence of A (L1)
A=/= B; B/C=A; B/C=/= B; I conclude negative influence of A (L 1)
C/D=/= A & B. I conclude that C & D behave in a SPECIFIC way different from both A & B.
I'm afraid, planned comparison would take much time, besides I might get confused cause I have many variables, besied I get used to ANOVA.
Rich Ulrich-2 wrote
Read a textbook!
(1) Apparently the ordering of the means is (B, A, C)
and only the extreme two show a difference.
That is a perfectly ordinary result for such testing.
There is nothing at all contradictory about it.
What is mysterious is why you say there are 4 groups but
name only A, B, and C.
(2) The G-H test might be overly liberal (from what I gather
just now, Googling), when group sizes are equal and small.
You report that it gives you slightly different results from
the other, but there is nothing wrong with the results from
the other test.
With two control groups, it sounds like you might be more
properly served by doing planned comparisons: If the two
control groups do not differ, then you carry out your *main*
test as the full-size, nominally 5% test comparing (A,B=controls)
to C. Here is a reference that discusses various post hoc tests
and recommends that planned comparisons are usually more
powerful and appropriate.
http://beheco.oxfordjournals.org/content/19/3/690.full--
Rich Ulrich
> Date: Sat, 29 Jun 2013 04:18:45 -0700
> From: [hidden email]> Subject: ANOVA-pos hoc
> To: [hidden email]>
> Safe my life!!
> my study includes 4 groups, 20 subjects each. ANOVA with Tukey/Bonferroni
> test shows seemingly contradictory results;
> 1. Though A and B (controlling groups) are equal (not significant), C-A is
> equal (insignificant), but C-B is not equal (significant).
> 2. I used Games-Howel post hoc since Levene's Test indicating unequal
> variance (less that 0.05), it really reduces the occurrence of such results,
> but still encountering few cases.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ANOVA-pos-hoc-tp5720943.html> Sent from the SPSSX Discussion mailing list archive at Nabble.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
Rich Ulrich-2 wrote
Read a textbook!
(1) Apparently the ordering of the means is (B, A, C)
and only the extreme two show a difference.
That is a perfectly ordinary result for such testing.
There is nothing at all contradictory about it.
What is mysterious is why you say there are 4 groups but
name only A, B, and C.
(2) The G-H test might be overly liberal (from what I gather
just now, Googling), when group sizes are equal and small.
You report that it gives you slightly different results from
the other, but there is nothing wrong with the results from
the other test.
With two control groups, it sounds like you might be more
properly served by doing planned comparisons: If the two
control groups do not differ, then you carry out your *main*
test as the full-size, nominally 5% test comparing (A,B=controls)
to C. Here is a reference that discusses various post hoc tests
and recommends that planned comparisons are usually more
powerful and appropriate.
http://beheco.oxfordjournals.org/content/19/3/690.full--
Rich Ulrich
> Date: Sat, 29 Jun 2013 04:18:45 -0700
> From: [hidden email]> Subject: ANOVA-pos hoc
> To: [hidden email]>
> Safe my life!!
> my study includes 4 groups, 20 subjects each. ANOVA with Tukey/Bonferroni
> test shows seemingly contradictory results;
> 1. Though A and B (controlling groups) are equal (not significant), C-A is
> equal (insignificant), but C-B is not equal (significant).
> 2. I used Games-Howel post hoc since Levene's Test indicating unequal
> variance (less that 0.05), it really reduces the occurrence of such results,
> but still encountering few cases.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ANOVA-pos-hoc-tp5720943.html> Sent from the SPSSX Discussion mailing list archive at Nabble.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
Rich Ulrich-2 wrote
Read a textbook!
(1) Apparently the ordering of the means is (B, A, C)
and only the extreme two show a difference.
That is a perfectly ordinary result for such testing.
There is nothing at all contradictory about it.
What is mysterious is why you say there are 4 groups but
name only A, B, and C.
(2) The G-H test might be overly liberal (from what I gather
just now, Googling), when group sizes are equal and small.
You report that it gives you slightly different results from
the other, but there is nothing wrong with the results from
the other test.
With two control groups, it sounds like you might be more
properly served by doing planned comparisons: If the two
control groups do not differ, then you carry out your *main*
test as the full-size, nominally 5% test comparing (A,B=controls)
to C. Here is a reference that discusses various post hoc tests
and recommends that planned comparisons are usually more
powerful and appropriate.
http://beheco.oxfordjournals.org/content/19/3/690.full--
Rich Ulrich
> Date: Sat, 29 Jun 2013 04:18:45 -0700
> From: [hidden email]> Subject: ANOVA-pos hoc
> To: [hidden email]>
> Safe my life!!
> my study includes 4 groups, 20 subjects each. ANOVA with Tukey/Bonferroni
> test shows seemingly contradictory results;
> 1. Though A and B (controlling groups) are equal (not significant), C-A is
> equal (insignificant), but C-B is not equal (significant).
> 2. I used Games-Howel post hoc since Levene's Test indicating unequal
> variance (less that 0.05), it really reduces the occurrence of such results,
> but still encountering few cases.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ANOVA-pos-hoc-tp5720943.html> Sent from the SPSSX Discussion mailing list archive at Nabble.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
Rich Ulrich-2 wrote
Read a textbook!
(1) Apparently the ordering of the means is (B, A, C)
and only the extreme two show a difference.
That is a perfectly ordinary result for such testing.
There is nothing at all contradictory about it.
What is mysterious is why you say there are 4 groups but
name only A, B, and C.
(2) The G-H test might be overly liberal (from what I gather
just now, Googling), when group sizes are equal and small.
You report that it gives you slightly different results from
the other, but there is nothing wrong with the results from
the other test.
With two control groups, it sounds like you might be more
properly served by doing planned comparisons: If the two
control groups do not differ, then you carry out your *main*
test as the full-size, nominally 5% test comparing (A,B=controls)
to C. Here is a reference that discusses various post hoc tests
and recommends that planned comparisons are usually more
powerful and appropriate.
http://beheco.oxfordjournals.org/content/19/3/690.full--
Rich Ulrich
> Date: Sat, 29 Jun 2013 04:18:45 -0700
> From: [hidden email]> Subject: ANOVA-pos hoc
> To: [hidden email]>
> Safe my life!!
> my study includes 4 groups, 20 subjects each. ANOVA with Tukey/Bonferroni
> test shows seemingly contradictory results;
> 1. Though A and B (controlling groups) are equal (not significant), C-A is
> equal (insignificant), but C-B is not equal (significant).
> 2. I used Games-Howel post hoc since Levene's Test indicating unequal
> variance (less that 0.05), it really reduces the occurrence of such results,
> but still encountering few cases.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ANOVA-pos-hoc-tp5720943.html> Sent from the SPSSX Discussion mailing list archive at Nabble.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
Rich Ulrich-2 wrote
Read a textbook!
(1) Apparently the ordering of the means is (B, A, C)
and only the extreme two show a difference.
That is a perfectly ordinary result for such testing.
There is nothing at all contradictory about it.
What is mysterious is why you say there are 4 groups but
name only A, B, and C.
(2) The G-H test might be overly liberal (from what I gather
just now, Googling), when group sizes are equal and small.
You report that it gives you slightly different results from
the other, but there is nothing wrong with the results from
the other test.
With two control groups, it sounds like you might be more
properly served by doing planned comparisons: If the two
control groups do not differ, then you carry out your *main*
test as the full-size, nominally 5% test comparing (A,B=controls)
to C. Here is a reference that discusses various post hoc tests
and recommends that planned comparisons are usually more
powerful and appropriate.
http://beheco.oxfordjournals.org/content/19/3/690.full--
Rich Ulrich
> Date: Sat, 29 Jun 2013 04:18:45 -0700
> From: [hidden email]> Subject: ANOVA-pos hoc
> To: [hidden email]>
> Safe my life!!
> my study includes 4 groups, 20 subjects each. ANOVA with Tukey/Bonferroni
> test shows seemingly contradictory results;
> 1. Though A and B (controlling groups) are equal (not significant), C-A is
> equal (insignificant), but C-B is not equal (significant).
> 2. I used Games-Howel post hoc since Levene's Test indicating unequal
> variance (less that 0.05), it really reduces the occurrence of such results,
> but still encountering few cases.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ANOVA-pos-hoc-tp5720943.html> Sent from the SPSSX Discussion mailing list archive at Nabble.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
Rich Ulrich-2 wrote
Read a textbook!
(1) Apparently the ordering of the means is (B, A, C)
and only the extreme two show a difference.
That is a perfectly ordinary result for such testing.
There is nothing at all contradictory about it.
What is mysterious is why you say there are 4 groups but
name only A, B, and C.
(2) The G-H test might be overly liberal (from what I gather
just now, Googling), when group sizes are equal and small.
You report that it gives you slightly different results from
the other, but there is nothing wrong with the results from
the other test.
With two control groups, it sounds like you might be more
properly served by doing planned comparisons: If the two
control groups do not differ, then you carry out your *main*
test as the full-size, nominally 5% test comparing (A,B=controls)
to C. Here is a reference that discusses various post hoc tests
and recommends that planned comparisons are usually more
powerful and appropriate.
http://beheco.oxfordjournals.org/content/19/3/690.full--
Rich Ulrich
> Date: Sat, 29 Jun 2013 04:18:45 -0700
> From: [hidden email]> Subject: ANOVA-pos hoc
> To: [hidden email]>
> Safe my life!!
> my study includes 4 groups, 20 subjects each. ANOVA with Tukey/Bonferroni
> test shows seemingly contradictory results;
> 1. Though A and B (controlling groups) are equal (not significant), C-A is
> equal (insignificant), but C-B is not equal (significant).
> 2. I used Games-Howel post hoc since Levene's Test indicating unequal
> variance (less that 0.05), it really reduces the occurrence of such results,
> but still encountering few cases.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ANOVA-pos-hoc-tp5720943.html> Sent from the SPSSX Discussion mailing list archive at Nabble.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
Rich Ulrich-2 wrote
Read a textbook!
(1) Apparently the ordering of the means is (B, A, C)
and only the extreme two show a difference.
That is a perfectly ordinary result for such testing.
There is nothing at all contradictory about it.
What is mysterious is why you say there are 4 groups but
name only A, B, and C.
(2) The G-H test might be overly liberal (from what I gather
just now, Googling), when group sizes are equal and small.
You report that it gives you slightly different results from
the other, but there is nothing wrong with the results from
the other test.
With two control groups, it sounds like you might be more
properly served by doing planned comparisons: If the two
control groups do not differ, then you carry out your *main*
test as the full-size, nominally 5% test comparing (A,B=controls)
to C. Here is a reference that discusses various post hoc tests
and recommends that planned comparisons are usually more
powerful and appropriate.
http://beheco.oxfordjournals.org/content/19/3/690.full--
Rich Ulrich
> Date: Sat, 29 Jun 2013 04:18:45 -0700
> From: [hidden email]> Subject: ANOVA-pos hoc
> To: [hidden email]>
> Safe my life!!
> my study includes 4 groups, 20 subjects each. ANOVA with Tukey/Bonferroni
> test shows seemingly contradictory results;
> 1. Though A and B (controlling groups) are equal (not significant), C-A is
> equal (insignificant), but C-B is not equal (significant).
> 2. I used Games-Howel post hoc since Levene's Test indicating unequal
> variance (less that 0.05), it really reduces the occurrence of such results,
> but still encountering few cases.
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ANOVA-pos-hoc-tp5720943.html> Sent from the SPSSX Discussion mailing list archive at Nabble.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
|
Locked
