0
$\begingroup$

I am analyzing data from different dichotomous variables. When analyzing association between these variables, what is the most recommended method?

$\endgroup$
  • 1
    $\begingroup$ There exist dozens of measures of association between dichotomous or binary variables. Note that in some situations binary (= "ordinal dichotomuus": present vs absent) should be treated differently than "nominal dichotomous" (eg male vs female) $\endgroup$ – ttnphns Jan 29 '13 at 9:36
  • $\begingroup$ @ttnphns: Could you please elaborate a little on ordinal dichotomous vs nominal dichotomous ? - sounds interesting. $\endgroup$ – Scortchi - Reinstate Monica Jan 29 '13 at 11:29
  • $\begingroup$ Agree with @ttnphns. I have mainly ordinal dichotomous in my study vs. some categorical variables. How can I find associations between dichotomous and categorical variables? $\endgroup$ – Wendy Alfaro Jan 29 '13 at 12:07
  • 1
    $\begingroup$ @Scortchi, at first glance all dichotomous attributes are formally the same. But if you compare logic behind some association indices, for example, Jaccard coefficient vs Russel-Rao coefficient, the difference becomes apparent. "Attribute X present vs attribute X absent" isn't quite the same as "Attribute X present vs alternative attribute Y present". $\endgroup$ – ttnphns Jan 29 '13 at 12:53
  • 2
    $\begingroup$ Of possible interest: Cross tabulation of two categorical variables: recommended techniques. If you can't give us more details on the purpose of the analysis and how data were collected, this question is no more than a duplicate. $\endgroup$ – chl Apr 29 '13 at 10:52
0
$\begingroup$

phi coefficient if you want a measure between -1 and 1. Chi-sq test or Fisher's exact test if you want to test for association.

$\endgroup$
  • 4
    $\begingroup$ We generally choose a statistical test or measure of association to answer a particular question given a sampling scheme or experimental design. $\endgroup$ – chl Apr 29 '13 at 10:49

Not the answer you're looking for? Browse other questions tagged or ask your own question.