Cochran and McNemar - quick help needed

[Charlotte-9]

Dear all,

Just a quick question.  I am comparing two proportions for the same group
of people.  E.g. One group of people, 65% undertook a test at one point in
time and 70% of those people undertook the same test at a later point in
time.  I want to find out if the increase was significant.

Is it correct to use McNemar or Cochran for this or could I use either?
If I understand correctly, they should give the same result as Cochran is
an extension of McNemar, but I want to ensure I'm using the appropriate
test for my report.

Thanks in advance,

Lou

[Marta García-Granero]

Hi Lou

Two related samples, McNemar test. Three or more related samples,
Cochran test.

Although you could use Cochran test for a two sample design (as you
could use an ANOVA instead of Student t-test), the most appropriate
test is McNemar. You could (and should) accompany your result with a
95%CI for the difference of the two proportions. Since SPSS doesn't
provide it (and I think it is about time SPSS started to focus more on
confidence intervals as standard output), you can find it here:

http://www.kingdouglas.com/SPSS/DiverseCultures/Marta/Code/BMJ%20-%20Stats%20Square%20One.txt

(Kindly hosted by King Douglas)

You will find MATRIX code to compute the 95%CI (asymptotic) right
after Table 6.1 dataset.

L> Just a quick question.  I am comparing two proportions for the same group
L> of people.  E.g. One group of people, 65% undertook a test at one point in
L> time and 70% of those people undertook the same test at a later point in
L> time.  I want to find out if the increase was significant.

L> Is it correct to use McNemar or Cochran for this or could I use either?
L> If I understand correctly, they should give the same result as Cochran is
L> an extension of McNemar, but I want to ensure I'm using the appropriate
L> test for my report.

--
Regards,
Dr. Marta García-Granero,PhD           mailto:[hidden email]
Statistician

(still recovering from a very bad flu+pharyngeal infection that kept
me far from the list)

---
"It is unwise to use a statistical procedure whose use one does
not understand. SPSS syntax guide cannot supply this knowledge, and it
is certainly no substitute for the basic understanding of statistics
and statistical thinking that is essential for the wise choice of
methods and the correct interpretation of their results".

(Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)

[cbautista]

Hi list,

I have a data on 500 injecting drug users and am planning to conduct a
cluster analysis in order to reveal groupings. This data have categorical
variables (i.e., gender, cleaning needles (yes/no), education (none,
primary, secondary, or higher...etc)) and quantitative variables (i.e.,
age, years of being an IDU, etc).

I was reading the literature on this topic (cluster analysis) and find that
most of cluster methods cann't use/include categorical variables. Besides,
I see that since the version 12.0 SPSS has the TwoStep Cluster Analysis,
where all type of variables can be used.

1) Is there any methodology to follow in the selection of cluster
variables?
2) Do you know pros and cons of the TwoStep Cluster analysis? I've found a
few articles where this method has been used.
3) In some published articles, the authors applied first, confirmatory
factor analysis to determine first-order factors and then they used these
first-order factors as the cluster variables. So, based on what conditions,
does this methodology has to be applied?

Thanks in advance,

/Christian