This is a huge area and this is going to be just a short summary. The shortest possible summary is probably that this is a regression analysis, i.e. it looks at whether, across a dataset, values on one or more variables appear to be systematically related to values on another variable. The key thing that distinguishes logistic regression from linear regression is that the predicted variable is binary (or categorical) not continuous and the relationship between the predictor and the predicted variable is about whether the predictor variable values appear related to the proportion(s) of the participants in one category of the dependent variable.
The analysis can tell you whether the relationship is statistically significant in the usual null hypothesis statistical testing (NHST) way, i.e. that the relationship was unlikely to have arisen by chance alone. It will also give you an index of the strength of the effect and can give you a confidence interval around that index. The analyses can handle multiple predictor variable or just one and, if there are more than one predictor, can explore interactions between predictors (i.e. that the effect of one predictor appears to be non-randomly related to the effect of another predictor).
Details #
That was the best I could do to describe logistic regression without so simplifying things that I misrepresent it. As with linear regression and correlation analyses, a statistically significant logistic regression doesn’t indicate that the relationship is causal. Equally, as with all NHST methods it’s crucial that the data are “independent” i.e. the likelihood of one participant being included is not systematically affected by another participant being included, so sampling everyone in a household is violating this requirement.
I think the best way I can explain a bit more about logistic regression is with some simulations or real data. However, that won’t fit in here so I will hope to expand on this in one or more Rblog posts.
Try also #
Chapters #
Not covered in the OMbook.
Online resources #
I hope to create at least one illustrative Rblog post.
Dates #
First created 12.ii.25.
