Hello
I had a question for you, please help me.
I know that beta regression is used when the response variable is relative.
Here my dependece variable is prevalence.
My independent variables are age (), season (1=Summer (reference)
2=winter 3=Spring 4=autumn), gender. , and so on.
My question is that when my independece variables have more than two classes or mode, (4 class ) same age and season.
y= prevalence
x1: age= category with 4 mode(1=10-20 years, 2= 20-30, 3= 30-40, 4= 40-50) or
x2: season (1=Summer (reference) 2=winter 3=Spring 4=autumn)
ello
x3: gender
How do I enter these variables (age and season) into the beta regression model? Do I have to enter them in a live form dummy? Or can I enter a category as a variable and use one of the modes as a reference?
Can we enter as a variable with 4 class and consider a class as a reference and measure the effect of beta (ceoffesion) on that reference?
What is the interpretation of B(coef) in beta regression?
Thank you for your response
thank you so much
But unfortunately there was no answer to my questions.
Hello Touba,
If the dependent variable is prevalence as a continuous variable, a Beta regression sounds right. That means any value (with infinite precision between a minimum and maximum are possible (e.g. 0% and 100%). If the prevalence is calculated from counts (e.g. 10 infected out of 1000), a Generalized Linear Model with Binomial distribution would be more appropriate.
* Independent variables with >2 classes: most modern software should be able to handle them as-is (and do the dummy coding internally). Just ensure that the categories are not treated as numeric and convert them if necessary.
* Reference class: Having one class as the reference is often the default (e.g. in R, the first class or level will be taken as reference - by default, the first class is determined with the alphabetical order but can be changed of course).
* Coefficients: the Beta regression has a link function (usually logit transform), so the coefficient is the slope of the linear association on the logit scale.
There are many resources online to learn more about beta regression (e.g. https://rcompanion.org/handbook/J_02.html)
Hope that helps
Thank you A reference class is placed to interpret the classification variable with 4 classes. If we have 4 classes with the codes 1, 2, 3 and 4 and the class 1 is a reference, is the interpretation of beta coefficients for classes 2, 3 and 4 relative to the reference, ie class one, measured? For example, age has 4 floors with classes 1, 2, 3 and 4. class 1 is a reference.To interpret the beta coefficient in class 2, we say: the average prevalence in the age group ( class 2) beta more or less than the reference age group (class one)? age: category variable with 4 class (cod:1,2,3,4)
Touba Narimani Moghadam ,
Yes, for a 4 class variable, the 3 coefficients will be the differences between the reference and each of the 3 other classes (on a logit scale)
I read in article (and help stata) that you should use marginal to determine the amount (Numerical value of B effect) of effect of beta coefficients.
Here is a link to the article that explains difference between beta and dirichlet regression: Article Analysing continuous proportions in ecology and evolution: A...
See figure 1 for an easy-to-read diagram and read it through the sample analysis.- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression