tag:blogger.com,1999:blog-3240271627873788873.post931905813029838332..comments2019-01-19T07:18:49.744-05:00Comments on Doing Bayesian Data Analysis: Don't treat ordinal data as metric -- update of movie ratingsJohn K. Kruschkehttp://www.blogger.com/profile/17323153789716653784noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-3240271627873788873.post-44551727177094295092018-03-23T10:16:57.824-04:002018-03-23T10:16:57.824-04:00When the independent variable is ordinal, you have...When the independent variable is ordinal, you have various options. Here are a few. One option is to treat the levels of the ordinal independent variable as categories of a nominal scale. That is, treat the IV as nominal (analogous to an ANOVA). That way you're predicting the DV for each level of the nominal IV and can do comparisons of levels of the IV, but you do not have any trend line across levels. As another option, you can pretend the levels of the ordinal IV are at estimated values on a continuous scale: Fix the first and last levels, and estimate the interior levels that best accommodate the trend you're assuming. Those two options assume you're doing regression of a DV on an IV. But another option is to treat both variables as DV's in a multivariate model. Then you're doing a sort of factor analysis, with a latent scale describing both ordinal values simultaneously.John K. Kruschkehttps://www.blogger.com/profile/17323153789716653784noreply@blogger.comtag:blogger.com,1999:blog-3240271627873788873.post-25673081971834333542018-03-23T00:54:36.946-04:002018-03-23T00:54:36.946-04:00I am currently helping a colleague make sense of h...I am currently helping a colleague make sense of her data. Both the outcome and covariate of main interest are Likert scales. Presumably the correct way of going about the analysis would be to first retrieve the underlying continuous construct for each variable, for each individual and then proceed with a standard linear regression between those inferred measures?<br /><br />Sorry if this is an obvious question but this is the first time I'm working (appropriately!) with this type of data. Really glad I found your preprint before mucking this one up.Manuelhttps://www.blogger.com/profile/17946463260666479941noreply@blogger.com