Categorical Data Analysis Lectures: Study Counts, Associations and Odds

Categories such as response type or treatment group produce counts rather than continuous measurements. Categorical data analysis lectures organize those counts into tables and examine conditional proportions, independence and logistic models. Following the denominator of each comparison is crucial: row percentages, column percentages, risks and odds can describe different aspects of the same table.

Here's a popular Categorical Data Analysis video — watch it, then summarize it with one click. Or paste any other YouTube link below.

— or paste your own YouTube link —

Want to keep the key points from a YouTube video? Paste the link below to turn it into structured notes.

Why Use Categorical Data Analysis Lectures: Study Counts, Associations and Odds

Frequently Asked Questions

Why can row and column percentages tell different stories?

They condition on different groups and therefore use different denominators. A row percentage describes composition within a row, while a column percentage describes composition within a column; neither should be interpreted without its conditioning variable.

What do expected counts mean in a chi-square independence test?

They are the counts implied by an independence model using the observed margins. They provide a comparison for the actual table, rather than predicting the precise counts that must occur in another sample.

Are odds and probability interchangeable?

Probability compares an event with all possibilities, while odds compare the event with its complement. Their numerical values can differ substantially, especially for common events, so an odds ratio should not be relabeled a risk ratio.

Why can sparse cells complicate a chi-square approximation?

The reference distribution is an approximation whose adequacy depends on the table and expected frequencies. With sparse information, an exact or carefully chosen alternative procedure may better reflect the discrete sampling structure.

What does a logistic regression model predict?

It models the log odds of a binary outcome as a function of predictors, which can be transformed into a probability. Interpretation of a coefficient depends on predictor coding and which other variables are included.