Question

a.

a.Your CMO wants to test the response to a new product that is designed to appeal to a subset of consumers - those that have lived in their current address for less than a year. She asks you to run a test and measure response based on a sample of 500 people. She suggests selecting the 500 consumers from the database in the following way: For each record, flip a coin. If the coin comes up Heads, then the consumer is included in the test. If Tails, then the consumer is not included. Is this a good way to assemble a sample to test response to the product? Why or why not?                                                                                                                                                                                                                                                        b.In a meeting, a Senior Manager says that a “negative correlation” doesn’t sound like a desirable result, and a “perfect correlation” sounds like a good thing. So he is confused about the meaning of a “perfectly negative correlation”. How would you respond?

Answer 1

A) no its not a good way. Using the probability of heads as 0.5, roughly half the people will be included in our sample space. Such a large sample size is infeasible and not necessary to generate meaningful conclusions about the population. Its not a good way

B) i would tell them that perfect correlation is just a way of saying the two things are strongly linked. Negative and positive are the direction of the linking. For example age is both perfectly and negatively correlated with chances of getting cancer. This is because young people very rarely get the disease and older people often do.

I would explain that perfect signals strength and negative signals direction. So 'perfect negative correlation' is a strong correlation where the second variable decreases when the first increases