dichotomous variables - statistical tests

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

dichotomous variables - statistical tests

Mark Lenel
Dear all,
 
Quick question which I'm hoping has an obvious answer to those in the
know!
 
I have a sample drawn from the population, and a sample statistic of a
dichotomous variable (e.g. 37% of people have a pet dog).  I then take a
subset of this sample, e.g. 18-24 year olds, and another statistic (e.g.
45% of 18-24 year olds have a pet dog).  I want to know if this second
statistic is significantly different from the first.  What is the
appropriate statistical test to use in this case?
 
I have my doubts about using a t-test because one of the samples is a
sub-set of the first.
 
Is it sufficient to calculate confidence limits around the total sample
statistic, and see if the subgroup statistic lies inside or outside
these limits?
 
Thanks,
Mark
Reply | Threaded
Open this post in threaded view
|

Re: dichotomous variables - statistical tests

Dominic Lusinchi
Mark,

Cross-classify (crosstab) age with pet ownership (both categorical
variables, one of which is ordered - age).

You can use the chi-square test to establish, in the first instance, whether
or age and pet ownership are independent: i.e., answer the question "Is
there an association between age and pet ownership?" Make sure that your
data set meets the test's requirements: no more than 20% of cells have
*expected* frequencies less than 5, and all cells have *expected*
frequencies of 1 or more. SPSS will tell you this at the bottom of the
chi-square tests table.

If you find that age and pet ownership are not independent (i.e. are
associated), then you can explore your contingency table further (partition
chi-square): this will help you determine which categories contribute most
to the association between age and pet ownership.

Cheers,
Dominic

Dominic Lusinchi
Statistician
Far West Research
Statistical Consulting
San Francisco, California
415-664-3032
www.farwestresearch.com
-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Mark Lenel
Sent: Friday, November 03, 2006 3:43 AM
To: [hidden email]
Subject: dichotomous variables - statistical tests

Dear all,

Quick question which I'm hoping has an obvious answer to those in the
know!

I have a sample drawn from the population, and a sample statistic of a
dichotomous variable (e.g. 37% of people have a pet dog).  I then take a
subset of this sample, e.g. 18-24 year olds, and another statistic (e.g.
45% of 18-24 year olds have a pet dog).  I want to know if this second
statistic is significantly different from the first.  What is the
appropriate statistical test to use in this case?

I have my doubts about using a t-test because one of the samples is a
sub-set of the first.

Is it sufficient to calculate confidence limits around the total sample
statistic, and see if the subgroup statistic lies inside or outside
these limits?

Thanks,
Mark