predictive analytics: explanatory

Question 1

predictive

Answer 1

• predict outcomes of Y given X
• what it means not as important
• models complex

Question 1

explanatory

Answer 1

• understand how Y is affected by X
• models simpler

Question 2

decisions informed by predictive analytics

Answer 2

• quality control; eg fraud detection, junkmail
• inventory management; sales forecasting
• risk analysis; churning, staff turnover
• market segmentation; who are least/most satisfied customers

Question 3

to estimate association

Answer 3

least squares estimates parameters

Question 4

to determine accuracy of coefficient estimates

Answer 4

confidence interval

Question 5

to test if there is a relationship between x and y

Answer 5

hypothesis test; null and alternative

Question 6

to determine how much model fits

Answer 6

R squared

Question 7

to determine a quantity

Answer 7

qualitative predictors

Question 8

professional reporting of results

Answer 8

• descriptive variable names
• avoid scientific notation
• avoid numerical clutter
• report p-values
• report sample size + measure of fit
• include mean of dependent variable
• table alternatives
• interpretation of other estimates
• facts to provide context
• description of modelling
• data cleaning and model selection decisions

Question 9

model building steps

Answer 9

• define business objectives
• collect/obtain data
• prepare/explore data
• create data for analysis/evaluation
• build/improve model
• deploy model

Question 10

build and improve model

Answer 10

• specify model
• estimate parameters
• interpret results and draw conclusions