Statistics lecture in workplace

WeeG

TS Contributor
#1
Hello

I have an unusual question. My boss suggested that I will give a lecture in statistics in my work place, where most of the people are scientists with no knowledge or basic knowledge in statistics.

I already did such a thing in the past, I gave a basic lecture covering basic topics like hypothesis testing rational, CI and so on. Now I need to be more specific and to focus on 1 applied topic. For example, something that can be relevant is quality control.

What I wanted to ask, is if any of you guys have other ideas for interesting lectures ?
The lecture should be approximately one hour and with minimum maths (although doesn't have to be no maths at all)

thanks !
 

hlsmith

Less is more. Stay pure. Stay poor.
#3
Nothing is better than taking an actual problem or current issue and using statistics to tackle it. Can you use an example of a previous Quality Improvement project that was conducted that most of the audience can related to?

Bigger question, how did Dason miss a post, what was he doing at 5:45 on the 28th of October?
 
#4
Well, I (for one) spent October 28th preparing an elaborate Halloween costume. So I'm going to give myself a pass!

In my experience talking to staff and managers that have a science background but no statistics background, I'm going to say start small. My challenge has been that despite their educational background, they tend to disbelieve the reliability of statistics. They have absolutely no doubt about the accuracy of sampling chemicals in a drum, but they inherently don't believe you can accurately sample human beings.

The way I've started to overcome that is to use Nate Silver's 538 info. Even if they don't know anything about statistics, my audience automatically gives a level of credibility to his work because they heard about it on TV. All you have to say is "correctly predicted 49 out of 50 states" and they listen, regardless of political affiliation. Most of what I do is related to public policy, so using Nate's work is familiar to my audience.

So my first recommended lecture topic would be something like: "Why is X survey/poll/something else you've heard of reliable (and how reliable is it really)?" Take something they already know a little bit about and explain how it works. Use the Presidential race or something else that is familiar to your audience.

If you are up for using 538's work, you can use it to do a talk on sampling and sampling bias ("if you don't get a good sample, you can't trust the results" seems to be a good lesson related to QC).

Doing someting about observational bias and the Hawthorne Effect would also be good QC topics.

I guess my overriding advice would be: don't focus on the math. It makes their eyes glaze over. Instead, focus on the practical outcomes that bad and good statistical methodology causes. Bad stats = unreliable results.

EDIT: Oops. Sorry. Now I see your location. OK, so 538 may not be common knowledge to you or your audience. But his work is still presented in an easy-to-digest format and might still be helpful. He's at http://fivethirtyeight.blogs.nytimes.com/.
 
Last edited: