1. Aprender
    2. /
  2. Cursos
    3. /
  3. Generalized Linear Models in Python
  • 1

    Introduction to GLMs

    Gratuito

    Review linear models and learn how GLMs are an extension of the linear model given different types of response variables. You will also learn the building blocks of GLMs and the technical process of fitting a GLM in Python.

  • 2

    Modeling Binary Data

    This chapter focuses on logistic regression. You'll learn about the structure of binary data, the logit link function, model fitting, as well as how to interpret model coefficients, model inference, and how to assess model performance.

  • 3

    Modeling Count Data

    Here you'll learn about Poisson regression, including the discussion on count data, Poisson distribution and the interpretation of the model fit. You'll also learn how to overcome problems with overdispersion. Finally, you'll get hands-on experience with the process of model visualization.

  • 4

    Multivariable Logistic Regression

    In this final chapter you'll learn how to increase the complexity of your model by adding more than one explanatory variable. You'll practice with the problem of multicollinearity, and with treating categorical and interaction terms in your model.

Initializing

Exercicio

Rate of change in probability

For the wells dataset you have already fitted a logistic regression model with the model formula switch ~ distance100 obtaining the following fit $$ log(\frac{\mu}{1-\mu}) = 0.6060 - 0.6219\times distance100 $$

In this exercise you will use that model to understand how the estimated probability changes at a certain value of distance100, say 1.5 as depicted in the figure below.

Recall the formulas for the inverse-logit (probability)

$$ \mu = \frac{exp(\beta_0+\beta_1x_1)}{1+exp(\beta_0+\beta_1x_1)} $$

and the slope of the tangent line of the model fit at point \(x\):

$$ \beta*\mu(1-\mu) $$

Dataset wells and the model wells_GLM are loaded in the workspace.

Instruções 1/3

undefined XP
    1
    2
    3
  • Define x to be the distance100 value of 1.5.
  • Extract model coefficients and save them as intercept and slope respectively.