MCQ: PICK THE RIGHT ANSWER:

ANSWER ALL THE ABOVE questions.

If we are doing missing data imputation, why might we consider a multi-layer neural network or k-nearest neighbours, instead of using linear regression? They require less data for training. They can learn more complex relationships between the features and the outcome variable. They are faster to train. They are easier to implement. If we have a binary classifier that always predicts 1 (the positive class), we will get: perfect precision perfect recall a recall of 0 a precision of 0 QUESTION 4 With capture-recapture techniques, our population estimator has several assumptions. If it turns out later that some identifier tags had been slipping off of fish between the first and second samples, describe which assumption is violated. No errors matching items across lists. Independence of samples. Homogeneous population. No outliers