How do they justify themselves, passing your opinion by? We don’t have time for a study? There are other factors besides stastical ones? There’s not enough data to form a basis for analysis?
In all these cases, it seems like the logical answer (which I realize they’ll wiggle out of) is: “I’m the statistician, if I can’t create a reliable analysis, I’ll tell you so–it’s not your job to assess a situation in my area of expertise.”
I took statistics in college. What impressed me is how the majority of people can’t work even with implications of simple figures. Do you intentionally “dumb down” results to make them more palatable? How?
The company I work for is a Six Sigma believer - and I’m a Six Sigma Brown Belt. Despite being math illiterate, I’ve had to get through an understanding of stats…and know exactly how they justify themselves…
“You need how many samples for that? We can’t do that”
“Yes, but there is still a chance you are wrong - that’s what you are telling us - and we like this answer.”
“Do you know how expensive that DOE would be - we can’t just change our machines like that?”
“We need answers now, we don’t have time to gather data and run an analysis.”
To get 3000 samples required to find a change in the defect rate could be cost prohibitive if the defect rate isn’t costing you that much to begin with, but it costs you $4 million to retool - or a complete plant shutdown and restart - which costs millions in lost productivity.
And sometimes companies spend far too much time and effort doing statistical analysis on really simple process problems - it takes time to gather and analyze data - it take very little time to realize that the reason your sales have dropped is that the store next door is selling everything you sell for 30% less - don’t need to do a lot of statistics there - yet you’d be surprised and the number of projects that we spend time gathering statisical data on when there is an obvious process change that could fix the issue in one quick step.
Well, sometimes relationships just run their course. A couple might date for a year, live together for three years, marry, then divorce after another eight years. They might also date for a year, marry, then divorce after eleven years. Same thing. (Yes, I know that’s not all there is to it; just saying that a lot of people marry, thinking they’ll be together forever, and it doesn’t happen. I’m not yet convinced that living together first dooms a marriage.)
And although this is not Cafe Society, I must post the Simpsons dialogue that came to mind while I was reading the OP:
Homer: Not a bear in sight. The Bear Patrol must be working like a charm.
Lisa: That’s spacious reasoning, Dad.
Homer: Thank you, dear.
Lisa: By your logic I could claim that this rock keeps tigers away.
I’m in the same boat. It’s a sad commentary that “The scientific truth” is rarely a match for “The attention-getting PR quote”.
The other pet peeve I’ve had is watching managers ask statisticians for a specific answer. “I want to say that this and that are related.” Well, the data says they’re not. “What if we look at it this way instead?” Still no. “How about this way?” Nope. (repeat until required conclusion is found)
You may be onto something there, Necros–since such liquids almost inevitably contain some amount of Dihydrogen Monoxide. This linked page includes a section entitled “FAQ: DHMO and Cancer”, which states:
Business systems programming works the same way. We have a saying to cover this situation (although we get ignored, anyway–management is generally not much concerned with reality):
– You WILL do the analysis. You can do it ahead of time and reduce costs by building it right, or you can do it during testing or after implementation when the costs of correction will be astronomical, but you WILL do the analysis for the system.
I guess then people teaching statistics should use that old rape/ice cream example. Idiots might get ice cream banned in the US and I like ice cream.(Am I assuming too much in thinking everybody has heard that rape/ice cream statistics example?)
