Warning #3211: Chi-square goodness of fit test with a zero data point

[a.bl...@gmail.com]

Hi,

I am performing a chi-square goodness of fit test for 4 individuals
with an observed frequency of a particular behaviour. For one
individual, this frequency is 0, which makes SPSS exclude the
datapoint (with the warning #3211) and only analyse the remaining
three, which decreases the significance of the data. Is there any way
to get around this within this test?

Thanks,
Anne

[Bruce Weaver]

If I follow, you have the number of times some behaviour occurred for
each of 4 individuals, and for one person, the count = 0. Is that
right? If so, are you really happy to treat the data as if each
behavioural episode within an individual is independent of all other
episodes (as is required for the chi-square goodness of fit test)?

What is the question you are trying to address?

--
Bruce Weaver
bwe...@lakeheadu.ca
www.angelfire.com/wv/bwhomedir

[Lorenzo Sandini]

Bruce,

I have been reading questions and answers in this group for a long long
time, and the hints and support you provide here are really really
precious. Looking for answers by studying the theory is certainly a
excellent way to learn, but as in my case, the load of work at the
hospital and at home makes this type of studying a really slow process.

Learning by the example may be the lazy man's solution, but in the case
of young doctors with limited knowledge of statistics, loads of clinical
work, small children at home, a quick and straightforward answer is
time-saving, if not life-saving.

I appreciate you stimulating us to think by ourselves, but with your
extended knowledge of statistics, I would appreciate even more a
straight answer from time to time. :)

(I am not the top-poster in this thread)

Gratefully and respectfully
Lorenzo Sandini
Kuopio, FINLAND

[Art Kendall]

As a physician you should be well aware of the distinction between the
presenting question and the underlying question.

Dr. Weaver sensibly asked the OP to clarify the question she was posing.
In statistical consulting as in many other fields such as nursing,
medicine, counseling, law, etc., an important skill is to be able to
move from a presenting question to a clearer underlying question. A
"straight" answer would have been meaningless and potentially deceptive.
It would have been of no use to the OP or to other readers of this news
group.

Art Kendall
Social Research Consultants

[Lorenzo Sandini]

Well Art, my limited knowledge of statistics in behavioural science does
not allow me to go forward. If the topic would be cigarette smoking for
example, the analysis would be even more intricated by the
inter-individual interactions, such as smoking in groups between
classes, etc... and I would request the help of a statistician.

What Bruce intented by saying that behavioural episodes in one
individual are related to other similar events in that same individual
is probably the key to solving this riddle in the statistically correct
way. If the outcome measure is an evaluation of addiction for example, I
would not know how to test the distribution of observed frequencies with
one of them being 0.

Again I am looking forward to reading your solution.

Lorenzo

[a.bl...@gmail.com]

Hi Bruce,

Thanks for your reply - see my follow-up questions interspersed with
your reply:

On Mar 29, 6:27 pm, Bruce Weaver <bwea...@lakeheadu.ca> wrote:

> a.blok...@gmail.com wrote:
> > Hi,
>
> > I am performing a chi-square goodness of fit test for 4 individuals
> > with an observed frequency of a particular behaviour. For one
> > individual, this frequency is 0, which makes SPSS exclude the
> > datapoint (with the warning #3211) and only analyse the remaining
> > three, which decreases the significance of the data. Is there any way
> > to get around this within this test?
>
> > Thanks,
> > Anne
>
> If I follow, you have the number of times some behaviour occurred for
> each of 4 individuals, and for one person, the count = 0. Is that
> right?

Yes, that's correct.

> If so, are you really happy to treat the data as if each
> behavioural episode within an individual is independent of all other
> episodes (as is required for the chi-square goodness of fit test)?

Yes, I am happy to treat the data as such.

>
> What is the question you are trying to address?

Essentially, can I perform a chi-square goodness of fit test when I
have a zero value for one frequency/count? As explained above, I get a
warning from SPSS, and would like to include this invididuals count in
the results, but at present that individual is excluded from the
results.

>
> --
> Bruce Weaver
> bwea...@lakeheadu.cawww.angelfire.com/wv/bwhomedir

Thanks,

Anne

[Bruce Weaver]

On Apr 3, 10:47 am, a.blok...@gmail.com wrote:

> > If I follow, you have the number of times some behaviour occurred for
> > each of 4 individuals, and for one person, the count = 0. Is that
> > right?
>
> Yes, that's correct.
>
> > If so, are you really happy to treat the data as if each
> > behavioural episode within an individual is independent of all other
> > episodes (as is required for the chi-square goodness of fit test)?
>
> Yes, I am happy to treat the data as such.
>
>
>
> > What is the question you are trying to address?
>
> Essentially, can I perform a chi-square goodness of fit test when I
> have a zero value for one frequency/count? As explained above, I get a
> warning from SPSS, and would like to include this invididuals count in
> the results, but at present that individual is excluded from the
> results.

You are running something like this:

NPAR TEST CHISQUARE= id
/EXPECTED=EQUAL
/MISSING ANALYSIS.

You need to change it to something like this:

NPAR TEST CHISQUARE = id(1,4)
/EXPECTED=EQUAL
/MISSING ANALYSIS.

The id(1,4) gives SPSS the lowest and highest values of ID.

[King]

On Apr 2, 7:51 am, Art Kendall <Arthur.Kend...@verizon.net> wrote:
> As a physician you should be well aware of the distinction between the
> presenting question and the underlying question.
>
> Dr. Weaver sensibly asked the OP to clarify the question she was posing.
> In statistical consulting as in many other fields such as nursing,
> medicine, counseling, law, etc., an important skill is to be able to
> move from a presenting question to a clearer underlying question. A
> "straight" answer would have been meaningless and potentially deceptive.
> It would have been of no use to the OP or to other readers of this news
> group.
>
> Art Kendall
> Social Research Consultants

Well stated, Art.

King Douglas
American Airlines Customer Research

[Bruce Weaver]

On 23/01/2016 6:02 PM, nourane...@gmail.com wrote:

> could you please tell me where exactly I can enter this command in SPSS? I have a similar problem and I can't find a way to code this in. I'm using SPSS 22, thanks.
>

You need to enter it in a syntax window. File > New > Syntax will open
a new syntax window. You can find many tutorials online on how to use
SPSS syntax. Here is one to get you started:

http://www.spss-tutorials.com/spss-syntax-editor-window/

HTH.

--
Bruce Weaver
bwe...@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/Home
"When all else fails, RTFM."