I've done much research, but it's difficult to get oriented without a specific background in statistics so any help would be appreciated. Since I'm the resident database/metrics/analysis professional I've been tapped to assist in answering some questions that HR leadership has presented. One such question is to try and prove or quantify a correlation between Base Salary, Salary Position in Range, and tenure and turnover.

I have two data sets that are identical except one is for employees who voluntarily left the company and the other is current employees in the same positions. The columns are Tenure (IN Months) / Annual Salary / PIR (As a %)


With some simple SQL I was able to derive some pretty obvious aggregate numbers which are displayed below (all #s are averages):


As you can see, people who leave are (outside of the first 2 tiers) generally paid much lower in their salary range and aren't as tenured. I have just recently begun a foray into R but am unsure as to which models/packages to use for multivariate testing.

The Ask: How can I visualize and/or model a statistical correlation between these variables in such a way I can say with a fair amount of confidence that they are related?

I also have access to pretty much every piece of comp or HR data imaginable so eventually I would then stress test these results against other variables, but to get started this would be amazing.

Does anyone have any pointers on which packages or models I should begin researching which would work best with the format of my data? Or pointers on a better way to arrange or structure the data in tandem with different packages (e.g. combining all data into 1 table with a Termination Flag). I'm not necessarily asking someone to walk me through it step by step, but a starting direction would be extremely helpful. I have already successfully utilized R to do some simple regression testing with ggscatter so I'm capable of using the program, just in some serious need of advice here.

  • $\begingroup$ Just use informative bar charts and use a logistic regression for the probability of leaving. $\endgroup$
    – AdamO
    May 31, 2018 at 19:20

1 Answer 1


You are on the right path. You have id codes, Jobs, with observations on four variables. The best way to start is to make a scatterplot matrix. Each of the six possible pairing will be displayed as a scatterplot with a point for each of the jobs. The r function plot in base R and plotmatrix in the ggplot2 package can generate these graphs. Also commercial statistical GUI software like JMP and SPSS supports this type of graphics.

However I don't understand the variable names and how they relate to the discussion. There are ways to get around naming conventions but the easiest thing to do is use'camel style' with first letters in a word capitalized i.e. BaseSalary

  • $\begingroup$ Thank you. I'll get started on trying to plot all this first and then go from there. Apologies on the variable names... this stuff comes out of business warehouse and I was just slapping it together with wanton disregard for naming conventions. $\endgroup$
    – Jesse H.
    May 31, 2018 at 19:37
  • $\begingroup$ Since I have the data broken up into two tables that are effectively identical in structure there's no way that I can think of to really relate them to the fact that some are terminations and some are not. Should I combine them with a binary column to flag terminations? If so, how does a binary value affect correlation? It seems like it wouldn't graph out very well since there's no range, it's either one or the other. $\endgroup$
    – Jesse H.
    May 31, 2018 at 19:57
  • $\begingroup$ Treat termination as a category. Stack the data and use different symbols to show current and terminated averages for each job. $\endgroup$
    – Georgette
    Jun 4, 2018 at 12:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.