Over 40 statistical analysis

Warning: geeky analysis coming up.

So those 40+ in age received a list of the ages of employees selected and not selected. Of course, I had to run some EDA and a hypothesis test. The comparative 5-number summaries and box-and-whisker plots were almost identical. I also ran a Wilcoxon rank sum test because of non-normal distributions. The results are as follows:

Wilcoxon rank sum test with continuity correction

data: age by status
W = 4330, p-value = 0.6751
alternative hypothesis: true location shift is not equal to 0

Given the p-value = 0.671, which is far greater than the null hypothesis at 0.05, we can reject the null and conclude that the distribution of age over selected and non-selected employees is the SAME.

TU did their homework.

Comment! It's anonymous!

8 replies (most recent on top)

How fun to get laid off and be part of a statistical model.

So I had a data friend do analysis as well. Just thought to get back to this. Their reaction to this post was

"oh good, I'm glad someone did something more serious with p-values and wrote it up. I definitely agree that it looks like TU did all things correctly to the T in aggregate, and also with the poster that there's not enough of a sample size to drill down too much."

From the data set I provided, there were 1552 people, 305 were RIFed - 19.6% of the department(s) provided.
mean age stay=44.1, rif=46.6
median age stay=44.0, rif=47.0

Thanks for answering my questions and your insights

In reply to @sck+1q4gF6PK. Plenty of young and old worker-bees cut. Cuts appear based on business segment and position, not age.

I asked that question.

So, they can still cut plenty of 40+ worker bees as long as they can keep 40+ management to balance the overall distribution and show that 40+ group is not disproportionately affected by layoffs.

In reply to @akp+1q4gF6PK, there is not a large enough sample size to segment age by category. In general, just eyeballing the data, it definitely appears DSA worker bees (Jr/Sr Analysts and Consultants) were hit the hardest. Curiously, no managers (I,II,III) were laid off in that group. Very few in data management, integration, and software development were hit. I was not provided stats for sales or other groups. My opinion is that analytic work has been or is in the process of being transferred to GCCs and/or Neustar, while managers, dev, and integration are still kept (at least for now) for the transition and any needed support/training for the replacements.

Thank you very much for doing this!

Can you control for the job level?

Curious to compare upper management vs worker bees within 40+ group.

author correction: should read "fail to reject the null hypothesis and conclude that we do not have sufficient evidence to say that the alternative hypothesis is true"

Conclusion remains.

my bad.

Thread regarding TransUnion layoffs

Over 40 statistical analysis

8 replies (most recent on top)

Post a reply