Explaining Analytics to a Non-Technical Manager - Tips, Examples, and Strategies

Common Mistakes and Misconceptions in Statistical Analysis

When asked to explain something to your manager, assume that person has a non-technical background and requires a brief but complete explanation of the issue.  Use this as an opportunity to demonstrate your thorough understanding of the material and your ability to communicate it effectively.

1.Your manager doesn’t know much about analytics.  That is too bad, but don’t worry, you’ll have his job soon enough.  In the meantime he has made the following comments.  If his suggestions / comments are good ones, explain why; if they are bad ones, explain why and what you should do to fix them or do better.  Make sure your answer demonstrates that you have a sophisticated understanding of the issues involved.

a.“If the p-value for the t-test is greater than alpha (say 0.06), the variable can’t belong in the model and should always be removed.”

b.“When you detect an outlier, you should just replace it with the mean value of the dataset.”

c.“When conducting a hypothesis test, like the t-test, you should always set alpha = 0.05.”

d.“Our hotel did a customer satisfaction survey for all the guests who visited last weekend.  The results were plagued by heteroskedasticity, so we basically had to scrap the results and start again.”

2.Collinearity: collections of variables that tend to move together, such as height and weight, are called collinear.  This creates some challenges for analysis in that individual t-statistics tend to be less informative.  Using the data found on the tab ‘Collinear’:

a.Filter the data to consider only the first 25 observations then run the following models; repeat the analysis with all the observations and note any differences.   You do not need to worry about standard data problems such as heteroscedasticity, etc. You will finish with 6 different regressions.

i.Run a linear regression to explain y in terms of experience and height.  Does height appear to explain y?

ii.Run a linear regression to explain y in terms of experience and weight.  Does weight appear to explain y?

iii.Run a linear regression to explain y in terms of experience and height and weight. Do height and weight appear to explain y?

b.Consider the results you suggest have found from the work in a. Write a paragraph or two to explain to your manager the patterns you observed with respect to the significance of the t-statistics, why these results occurred, and the strategies you for using explanatory variables that exhibit collinearity.