Logistic Regression Flashcards
What does a Logistic Regression imply?
That the outcomes are not numerical but categorical
Give an example using Linear Regression and Logistic Regression.
Linear - we can predict how much a customer will pay, if they buy
Logistic - we can predict whether a customer will buy at all
What does the logistic model do?
The logistic regression predicts the probability of an event occuring
What does MLE stand for?
Maximum Likelihood Estimation
What is the Log-likelihood?
the value is almost but not always negative
What is LLR?
Log Likelihood Ratio test - similar to F-stat - it measures if the model is useful
shown as a P-value
What is the Pseudo R-squared?
McFadden’s R-squared
A good value is between 0.2 and 0.4
used for comparing variations of the same model.
different models with have completely different and incomparable Pseudo R-squares
What are characteristics of underfitted models?
Low accuracy
They don’t predict well
What are characteristics of an overfitted model?
High train accuracy
what to do?
split dataset into train and test
What is the opposite of Accuracy?
The misclassification rate?
of misclassified / # of all elements
From which dataset do we judge the accuracy of a model? Train or Test?
Test
