Know-How Exchange : Research/Metrics

Dear Omar,

To the average employee, a term like "coefficient correlation"
is at best, mumbo jumbo, and at worst, meaningless BS.

For all the good it will do you, and where it really matters, you might as well be speaking Aramaic.

The problem with Pearson's calculations is they rely on relationships that are linear, and between points A and B.

But the workplace doesn't always work according to neat little sets of numbers: workplaces are organic or hierarchical, top down systems or middle out. Or so it seems to me.

Might it be better to put away your slide rule, to forget your calculations, and get out onto the shop floor, the production floor, or whatever the work environment you're studying is and TALK TO PEOPLE?

Employees are not mathematical units that you can assess and calculate, they're people.

They have homes and lives, problems and concerns, a sense of humour, issues, emotions, skills, baggage, experience, and memories and a good many other things ALL of which they bring
to work with them in one way or another.

I can pretty much guarantee that an older employee with more experience and on a lower pay grade, someone who's been doing the job and doing it well for years will resent a younger, higher paid employee either being promoted, or giving orders, or not "getting" it telling them what to do.

likewise, I can guarantee that a younger employee with a higher level of education than what they may see as someone who's less well educated and therefore (in the eyes of the younger person) not as intelligent resenting being given advice from someone they see as an old fart.

The same rule applies to skills, experience, perceived notions of the way things have been done and of the ways things need to be done and basically, your coefficient means nothing if people cannot get on, or if they will not take instructions or orders from someone they see as less than them in one way or another.

I've seen it happen and I've been on the receiving end of it.

The thing to do is to remove any and all assumptions—all of them, from the game. Once you do this, anything is possible and suddenly, any barriers that you might have had vanish.

If you attempt to point the blame for your numbers at any one person, if you go after any single employee, you begin a witch hunt; you, in essence, begin looking for a scapegoat when the real issue is poor management.

When you begin looking for culprits you kill morale, which means any numbers you run become even lower, which means you then look for more people to blame, so morale dips again, and so on.

Not good.

The answer to your question (or at least, MY answer) needs no calculation, no correlation of synergistic value adds, or any other convoluted description.

The answer is personal attitude and recognition for doing a great job.

I don't doubt that Karl Pearson was a brilliant mathematician,
but having done some reading on his methods and personal philosophies (he liked to apply social Darwinism to entire nations and, among other things he openly advocated what he termed "war" against "inferior races"), he wasn't exactly what one might call warm and fuzzy. His staunch support of eugenics might today label him as a right wing fundamentalist.

Does this kind of attitude or philosophy belong in any workplace?
I don't think so. You may not agree with my opinion, but it's what I think.

Don't measure people, don't compartmentalize them. Talk to them and find out what it is that drives them and that holds them back.
Most people WANT to do a good job. Give them the tools, the rewards, and the recognition they need and they'll reward you with loyalty.

Try to put people in boxes or try to apportion blame for some real or imagined short coming that management ought to have spotted and done something about and you'll create a monster.

I hope this helps.

Gary Bloomer
Wilmington, DE, USA

Hi Omar,

You are trying to detect an outlier, which can be done in several ways:
1) plot age and service years in a graph and eyeball a 'regression' line through the data (some statistical packages will do this for you), you can identify the outlier as the one far away from that line
2) for general statistical procedures, see pages 187-191 of RA Johnson, DW Wichern, 1998, Applied Multivariate Statistical analysis
3) in your case, with only 11 observations, you can simply use the 'leave one out' method: leave one observation out and calculate the correlation of the remaining 10. Repeat for each observation to leave out. The one left out that gives you a high correlation for the other 10 is the one responsible for the low correlation

Cheers