Date: Mon, 24 Jan 2011 12:45:18 -0500
Reply-To: R B <ryan.andrew.black@gmail.com>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: R B <ryan.andrew.black@gmail.com>
Subject: Re: Chi square or U Test
In-Reply-To: <1295879291934-3354708.post@n5.nabble.com>
Content-Type: text/plain; charset=windows-1252
Hi, Bruce:
In general, I think the approach you've suggested should work. I don't
have time right now to read the simulation study to which you linked
in the previous post. I will concur, however, that the abstract seems
to support use of a GLM. I guess my primary concern would be the
extent to which statistical power is compromised by how well the
assumptions of the specificifed GLM-ANOVA are met.
Ryan
On Mon, Jan 24, 2011 at 9:28 AM, Bruce Weaver <bruce.weaver@hotmail.com> wrote:
> Hi Ryan. If my description is correct, a two-factor (Phase x Category)
> repeated measures ANOVA would probably yield a reasonably good model too, I
> think. Ray Koopman, whom you know from his postings on other lists, once
> pointed to a paper by Shaffer (1981, "The analysis of variance mixed model
> with allocated observations") for situations like this, where the k-th score
> within a subject is perfectly predicted given the first k-1 scores. (I.e.,
> here, the 3 scores within a phase must sum to 20.) I see there is a more
> recent simulation paper that also concludes ANOVA works well for this
> situation.
>
> http://psycnet.apa.org/?fa=main.doiLanding&doi=10.1037/1082-989X.2.2.200
>
> HTH.
>
>
>
> R B wrote:
>>
>> Anthony,
>>
>> If Bruce's description of your data is correct, then you should
>> consider fitting a generalized logit model. Why? Because your response
>> consists of unordered categories. But we're not done! Not only are you
>> dealing with a multinomial distribution, but there is likely
>> covariation due to repeated measures. That is, you have (1) collected
>> multiple responses from the same person within Phase I, and if I
>> understand correctly, (2) these same individuals were measured again
>> during Phase II.
>>
>> A generalized logit MIXED model with appropriately specified fixed and
>> random effects can be fit employing the GENLINMIXED procedure in SPSS
>> 19.
>>
>> Ryan
>>
>> On Sun, Jan 23, 2011 at 9:01 AM, Bruce Weaver <bruce.weaver@hotmail.com>
>> wrote:
>>> Anthony, I see you're posting the same question under a new subject. I
>>> made
>>> a guess as to the structure of your data below, but did not see a
>>> response.
>>> Note that if you responded to my hotmail address, it is used only for
>>> posting to this list via Nabble, and I don't check it regularly--see the
>>> note in my signature file. So please respond to the mailing list.
>>>
>>> Bruce
>>>
>>>
>>> Bruce Weaver wrote:
>>>>
>>>>
>>>> Anthony James wrote:
>>>>>
>>>>> Dear Colleagues,
>>>>> I have a research design problem. I'd be thankful if you helped.
>>>>> I read 20 words for 30 cases and asked them to tell me the first word
>>>>> that hits them. I categorized the words generated by the participants
>>>>> under 3 categories of 'related words', 'irrelevant words' and
>>>>> 'synonyms/antonyms'. Then I computed the mean of each category over the
>>>>> 30 cases.
>>>>>
>>>>
>>>> You say you asked the cases to tell you the first word that hit them. I
>>>> take it, then, that cases are subjects (or participants) in your study,
>>>> right?
>>>>
>>>> Re the means you computed, I think they are the mean number of subjects
>>>> (or cases) who selected each category, is that right?
>>>>
>>>>
>>>> Anthony James wrote:
>>>>>
>>>>> In Phase 2, I give pictures as prompts to participants instead of words
>>>>> and asked them to tell me the first word that hits them. I categorized
>>>>> the words in the same way and computed the means of the three
>>>>> categories.
>>>>>
>>>>
>>>> Does Phase 2 use the same subjects (cases) in Phase 1?
>>>>
>>>>
>>>>
>>>> Anthony James wrote:
>>>>>
>>>>> Results show that the mean of 'synonyms/antonyms' when words are used
>>>>> as
>>>>> prompts is 19 and its mean is 13 when pictures are used as prompts.
>>>>> This
>>>>> suggests that words generate more synonyms/antonyms than pictures. How
>>>>> can I test the statistical significance of this difference? Is
>>>>> Mann-Whitney U Test the right test?
>>>>>
>>>>
>>>> The Mann-Whitney U test is a rank-based alternative to the unpaired (or
>>>> independent groups) t-test. I.e., it can be used when you have two
>>>> independent groups of observations.
>>>>
>>>>
>>>> Anthony James wrote:
>>>>>
>>>>> I can consider all the 3 categories of words and the two categories of
>>>>> picture and run a 2*3 chi square. However chi square doesn’t give the
>>>>> significance of the difference for each category separately.
>>>>> Â
>>>>> Cheers
>>>>> Anthony
>>>>> Â
>>>>>
>>>>
>>>> If you have the same subjects in both phases, then a chi-square test of
>>>> association is out--it requires independence of observations.
>>>>
>>>> I'll have a guess at what your data look like.
>>>>
>>>> Subject Phase Word Category
>>>> 1 1 1 2
>>>> 1 1 2 3
>>>> 1 1 3 1
>>>> 1 1 4 2
>>>> etc
>>>> 1 1 20 3
>>>> 2 1 1 1
>>>> 2 1 2 3
>>>> etc
>>>> 2 1 20 1
>>>> etc, including same type of thing with Phase = 2
>>>>
>>>> This can be reduced to one row per Subject as follows:
>>>>
>>>> Subject P1C1 P1C2 P1C3 P2C1 P2C2 P2C3
>>>> 1 8 7 5 2 10 8
>>>> 2 4 12 4 4 9 7
>>>> 3 11 3 6 5 6 9
>>>> etc
>>>>
>>>> Where P1C1 = Phase 1, Category 1; P1C2 = Phase 1, Category 2, etc; and
>>>> the
>>>> data points are counts of how often that category was chosen for each
>>>> subject. Thus, within a phase, C1, C2, and C3 must sum to 20.
>>>>
>>>> Is this what your data look like?
>>>>
>>>>
>>>>
>>>
>>>
>>> -----
>>> --
>>> Bruce Weaver
>>> bweaver@lakeheadu.ca
>>> http://sites.google.com/a/lakeheadu.ca/bweaver/
>>>
>>> "When all else fails, RTFM."
>>>
>>> NOTE: My Hotmail account is not monitored regularly.
>>> To send me an e-mail, please use the address shown above.
>>>
>>> --
>>> View this message in context:
>>> http://spssx-discussion.1045642.n5.nabble.com/Chi-square-or-U-Test-tp3347510p3353647.html
>>> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>>>
>>> =====================
>>> 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
>>
>>
>
>
> -----
> --
> Bruce Weaver
> bweaver@lakeheadu.ca
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Chi-square-or-U-Test-tp3347510p3354708.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> 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
| 
