one way ANOVA shows highly significant F ratio for 4 treatment group but duncan post hoc test does not produce significant at 0.01 level among few treatment

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

one way ANOVA shows highly significant F ratio for 4 treatment group but duncan post hoc test does not produce significant at 0.01 level among few treatment

mvishal8
This post was updated on .
In one way ANOVA, I got highly significant F ratio (p=0.00001) but when I ran duncan at 0.01 level , I didnot get significant difference in two treatment , but at 0.05 level I got difference in those treatment also. total 4 treatment comparision (including 1 control). so at what level I should declare my result...
Reply | Threaded
Open this post in threaded view
|

Re: one way ANOVA shows highly significant F ratio for 4 treatment group but duncan post hoc test does not produce significant at 0.01 level among few treatment

Maguin, Eugene
I think you are missing an important piece of information. With four groups (I'm assuming no control since you don't mention it), you can make 4*3/2=6 pairwise comparisons. The Duncan test has a specific procedure for ensuring that none of the possible comparisons will be significant if the null hypothesis is true. Take a little internet cruise through the topic of post hoc tests.
Gene Maguin

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of mvishal8
Sent: Tuesday, July 15, 2014 12:49 AM
To: [hidden email]
Subject: one way ANOVA shows highly significant F ratio for 4 treatment group but duncan post hoc test does not produce significant at 0.01 level among few treatment

In one way ANOVA, I got highly significant F ratio but when I ran duncan at
0.01 level , I didnot get difference in two treatment but at 0.05 level I got difference in that treatment also.



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/one-way-ANOVA-shows-highly-significant-F-ratio-for-4-treatment-group-but-duncan-post-hoc-test-does-nt-tp5726708.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

=====================
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
|

Re: one way ANOVA shows highly significant F ratio for 4 treatment group but duncan post hoc test does not produce significant at 0.01 level among few treatment

Rich Ulrich
In reply to this post by mvishal8
You do not ask a question.  I suppose that the question you intended
to ask would be along the lines of, "Is this reasonable?"  The short
answer is Yes, given the short description.

The Duncan test uses the studentized range for its test, which says
that three different cutoffs are used for the 5% (or 1%) test when you
compare the smallest mean with the 2nd ranked, 3rd ranked, and 4th ranked.
This makes the Duncan more conservative than the LSD test.  Since it uses a
single cutoff, LSD is (thus) easier to understand.  LSD also fits readily into a
testing narrative; I don't know what justifies using Duncan.

Almost all post-hoc tests where developed with the assumption that group
sizes are equal.  If yours differ, that is a possible complication with your
data and/or how SPSS handles this particular test.  I think IBM makes the
test algorithms available.

--
Rich Ulrich

> Date: Mon, 14 Jul 2014 21:48:33 -0700
> From: [hidden email]
> Subject: one way ANOVA shows highly significant F ratio for 4 treatment group but duncan post hoc test does not produce significant at 0.01 level among few treatment
> To: [hidden email]
>
> In one way ANOVA, I got highly significant F ratio but when I ran duncan at
> 0.01 level , I didnot get difference in two treatment but at 0.05 level I
> got difference in that treatment also.
>
>
===================== 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