albertorodrigues / generalized_linear_models

Generalized Linear Models

GLMs are a class of statistical models where there is more modeling possibilities than a simple linear regression.

There are many advantages to using them, some are:

1. Greater reliability in results
2. Greater Modeling Possibilities
3. Improved predictions

The big difference in this class is the possibilities of the probability distributions for the target variable.

The most common distributions are:

1. Normal
2. Inverse Normal
3. Gamma
4. Poisson
5. Negative Binomial
6. Binomial

Each model has a specific type of variable. For continuous data, the Normal, Gamma and Inverse Normal distributions are the most common alternatives while for count data or binary data the Poisson, Binomial and Negative Binomial distributions are the most common.

This project presents several real data analysis with this type of modeling. Including:

1. Creation of GLMs
2. Interpretation of results
3. creation of .confidence intervals
4. Comparison between models
5. Hypothesis testing for model parameters.

The data analyzes of this project were separated into three folders:

The first contains data analysis of data from the binomial distribution in which we have a certain number of successes and failures in the target variable.

Model creations were performed with various link functions and interpreting the model.

For further details, click here to go to these data analysis.

The second is in relation to classification models, specifically models in which the target variable has the bernoulli distribution, including logistic regression.

Several models were created with some data and predictions based on them. Also, we compare the prediction curves.

For further details, click here to go to these data analysis.

The last part concerns continuous and discrete data, where there is much more modeling possibilities.

Data analysis compared some models for this type of data, exploratory analysis and creation of confidence intervals.

For further details, click here to go to these data analysis.