/blog/2021/11/08/beta-regression-guide/

[andrewheiss]

A guide to modeling proportions with Bayesian beta and zero-inflated beta regression models | Andrew Heiss
https://www.andrewheiss.com/blog/2021/11/08/beta-regression-guide/

[andrewheiss]

Originally posted by 小牧和哉（KOMAKI Kazuya） on 2021-11-27 07:35:00

Thanks for the great post. (I also found your previous post on how to
calculate IPW very helpful.)

I have a few questions.

1. Why does the mean of the draws of the posterior distribution of
"Estimated marginal means" in "emeans" become "Average Marginal
Effects" (i.e., "Marginal effects are partial Marginal effects are
partial derivatives of the regression equation with respect to each
variable in the model for each unit in the data")?

２．In the calculation of interaction effects between categorical
variables（Include multiple categories）, "Pairwise comparisons of
average marginal effects" (which corresponds to "pwcompare" in Stata)
are obtained by the following method? (If polyarchy and corruption are
hypothetically considered as factors)

ame_fancy_zi_quota <- polyarchy + corruption %>%.
+ emmeans(~polyarchy|corruption, transform = "response") %>%
+ contrast(method = "revpairwise") %>%
+ gather_emmeans_draws()\

[andrewheiss]

Originally posted by Sebastian  on 2022-11-10 20:58:38

Hello, thank you very much for the very good and helpful post. I have a
question regarding fractional logistic regression:

I have a dataset on the proportion of inner urban development at the
level of municipalities (values from 0 to 1).

Now I want to estimate the proportion of inner urban development using
other independent variables. Since the dependent variable can only have
values from 0 to 1, fractional logistic regression is a good choice -
according to your post.

In your post you explain how to estimate the marginal effects of more
than one independent variable by holding the other independent variables
constant. In your example you do this for quota (binary) and for
polycratic (index from 0 to 1). This is very helpful. However, I need
help on how to calculate the marginal effects when you have four or
maybe more continuous independent variables (for example degree of
urbanization [values from 0 to 1], urban dispersion [index with
values from 0 to 100], population growth [percentage change],
building density [dwelling per built-up area]).

Can you give me advise how to calculate marginal effects in this case?

Best regards
Sebastian