Exam PA With Correct Solutions,100%

Exam PA With Correct
Solutions,100%.




Disadvantages of PCA - CORRECT ANSWER✔✔1) Some information will be lost
j j j j j j j j j j




2) PCs are very hard to interpret
j j j j j j




3) Considering the nature of the client's work, you may have to collect the original variables' values anyway, so
j j j j j j j j j j j j j j j j j j




it wouldn't actually save time.
j j j j j




Advantages of PCA - CORRECT ANSWER✔✔1) It can allow us to visualize high-dimensional data
j j j j j j j j j j j j j




2) It can make the model less complex by allowing it to run off only a few features
j j j j j j j j j j j j j j j j j

,3) It can identify latent variables (such as "overall health"
j j j j j j j j j




Principal Components Analysis - CORRECT ANSWER✔✔Summarizes high dimensional numeric data with
j j j j j j j j j j




fewer dimensions, while preserving the spread of the data. Can be particularly helpful with highly correlated
j j j j j j j j j j j j j j j j




variables. Need to also somehow mention that input variables should be centered and scaled
j j j j j j j j j j j j j j




Gamma Distribution - CORRECT ANSWER✔✔Non-negative and continuous
j j j j j j




Binomial Distribution - CORRECT ANSWER✔✔Works when there are only two outcomes
j j j j j j j j j j




Disadvantages of LASSO - CORRECT ANSWER✔✔"will eliminate so many variables that it predicts the same
j j j j j j j j j j j j j j




thing for someone who is ____ as someone who is ____"
j j j j j j j j j j j




Bias - CORRECT ANSWER✔✔expected loss caused by the model not being complex enough to capture the
j j j j j j j j j j j j j j j




signal in the data
j j j j




Variance (wrt bias-variance) - CORRECT ANSWER✔✔is the expected loss from the model being too complex
j j j j j j j j j j j j j j




and overfitting to the training data
j j j j j j




Total expected loss - CORRECT ANSWER✔✔Bias + Variance + Unavoidable error
j j j j j j j j j j




Disadvantages of GLM over Decision Tree - CORRECT ANSWER✔✔1) GLMs do not capture the effects of
j j j j j j j j j j j j j j j




variable interaction automatically. They have to be explicitly coded (must be put in business context)
j j j j j j j j j j j j j j j




Advantages of GLM over Decision Tree - CORRECT ANSWER✔✔1) In instances where a numeric variable
j j j j j j j j j j j j j j




simply is directly correlated with the response variable, trees would have to create a LOT of leaves to show
j j j j j j j j j j j j j j j j j j j




this.
j

, Advantages of GLM vs LASSO - CORRECT ANSWER✔✔1) LASSO "will eliminate so many variables that it
j j j j j j j j j j j j j j j




predicts the same thing for someone who is ____ as someone who is ____"
j j j j j j j j j j j j j j




2) Sometimes retaining seemingly insignificant variables can help with interpretation
j j j j j j j j j




Disadvantages of GLM vs LASSO - CORRECT ANSWER✔✔1) Overfitting j j j j j j j j




Create scatter plot - CORRECT ANSWER✔✔Geom_point()
j j j j j




Line graph - CORRECT ANSWER✔✔Geom_line()
j j j j




Imputation - CORRECT ANSWER✔✔Predict a missing value based on other variables
j j j j j j j j j j




How to justify that your model is useful - CORRECT ANSWER✔✔Compare to intercept only model that
j j j j j j j j j j j j j j j




predicts only the mean
j j j j




4 types of relationships to examine during exploration - CORRECT ANSWER✔✔1) Obvious relationship
j j j j j j j j j j j j




(premium vs coverage)
j j j




2) with response variable
j j j




3) those that surround potential hypotheses
j j j j j




4) Sense checking
j j




When to use log transform - CORRECT ANSWER✔✔Data are skewed, so transform in exploration stage, but
j j j j j j j j j j j j j j j




the response variable should NOT be transformed in the ultimate model
j j j j j j j j j j j




Leverage - CORRECT ANSWER✔✔How far a covariate x is from the other covariates
j j j j j j j j j j j j