Slashdot Mirror
Ask Slashdot: Resources For Explaining Statistics For the Very First Time? (thejuliagroup.com)
theodp writes: Teaching multivariate statistics to college students, writes AnnMaria De Mars, was a piece of cake compared to her current project — making a game to teach statistics to middle school students who have never been exposed to the idea. In the interest of making a better game, De Mars asks, "Here's my question to you, oh reader people, what resources have you found useful for teaching statistics? I mean, resources you have really watched or used and thought, 'Hey, this would be great for teaching?' There is a lot of mediocre, boring stuff on the interwebz and if any of you could point me to what you think rises above the rest, I'd be super appreciative." Larry Gonick's The Cartoon Guide to Statistics is pretty amazing, but is it a little too advanced for this age group? Anyone have experience with the Khan Academy Data and Statistics offerings? Any other ideas?
Khan Academy is probably the first thing I would choose.
An election year is always a good year to look at statistics because Nate Silver is always looking for trends.
Basic Statistics, you could use sports easily. Take the local high school teams and compare how they do to others and what their ratio of hits or misses are for their sport of choice.
Use their own past history in courses to determine how they will do in future courses. A history professor was pointing out that of those students in her class, those who looked at a specific resource had performed the best.
Using something that relates to the students in their day to day lives or at least something they find interesting to teach statistics will most likely yield the best results.
Place something witty here
As somebody who likes to teach math privately to people I recommend one thing first and foremost: Intuition. In mathematics, intuition is often thrown under the carpet as distracting from playing with mathematical concepts but in order to understand mathematics, you need to understand WHY people made formulas the way they do. As a result, students often have a 'see monkey, do monkey' mentality while having no true understanding of the topic. People with even less understanding aren't even able to replicate the desired results.
In general, the less the student has a feeling for mathematics, the more you need to teach intuition first and formulas later. Math students are of course required to have a higher level of understand, but this is obvious.
Knowledge is power. Knowledge shared is power lost.
What is there in statistics to make someone of that age care about it?
Teach them some dice based game, get them to play each other and mix in some loaded dice at random.
Tell them half way through and get them to figure out who is the cheat. Naturally they won't get it right as intuition about statistics is usually poor, but your job is to guide them into the right direction.
SJW n. One who posts facts.
You can add "how many team members are lying about not eating the M&Ms."
I think the problem is that statistics is a far more detailed and complicated topic than that, and that the sort of thing you suggested is the sort of thing that's already taught well, but such a tiny miniscule tip of the iceberg that it's the rest of it that needs be taught.
I agree with the person below who mentioned intuition, mostly the biggest problem I see when it comes to statistics amongst people of every age and group is that very few people seem to grasp the issues that may face a statistical result. People in general struggle to understand what the numbers actually mean, they're hopeless at figuring out what confounding factors may exist in a result.
So might I suggest a decent idea might be to find some bad statistical studies and create some exercises that help them understand why they're bad. The examples don't need to be difficult, but should be varied to help them understand why correlation does not mean causation, and why causation doesn't even necessarily imply (at least linear) correlation amongst other things.
Wikipedia's list would probably provide a reasonable starting point for some examples to cover:
https://en.wikipedia.org/wiki/...
I'd suggest, that by teaching kids how to question statistics, and how to spot when someone is using statistics to spout bullshit (which you'll find happens all the fucking time, basically every single day of your life if you're adept at spotting it) they'll be better placed to learn how to do statistics properly.
If they know how to tell when a result is wrong, they'll hopefully be encouraged to find out how to do it right, and how to mitigate these issues. I think without ever learning to recognise how people do statistics wrong on a daily basis, it's hard to know how to do it right, and so the issue just proliferates.
Lies, damned lies, and statistics.
https://en.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics
Statistics is best learned using a "Hands On" approach. It is a difficult subject for middle school students. An example lesson is to ask a relevant class question and then use the class data to teach what ever the topic is.
The professors of the California Math Project have access to a variety of practices and resources for teaching middle school math.
Try these resources; the National Counsel of Teachers of Mathematics,(NCTM), the Illustrative Mathematics web site, and "The Teaching Channel". These web sites have teaching activities and resources for middle school math
Here, teach them this, they should find this very interesting:
Imaginary Number Probability in Bayesian-type Inference
http://www.ccsenet.org/journal...
Everything I write is lies, read between the lines.
... here is the next book you need : How to Lie with Statistics ;)
http://www.amazon.com/How-Lie-...
Will $CURRENT_YEAR be the year of the Linux Desktop?
https://en.wikipedia.org/wiki/How_to_Lie_with_Statistics
Seriously - if you want to teach intuition with statistical models - and why most media published statistics are horribly developed - this is the book. The principle bits you get out of the book: Correlation does not imply causation - a HUGE intuitive bit of knowledge to debunk a LOT of what the media throws out there.
You know, 99.3% of all convicted criminals, in one form or another, have eaten a tomato! (Yes, ketchup and pizza sauce count)
100% of everyone who as ever, or will eat on, will die. (Note: did not say how long it would take for them to die!)
How do your correlate your data so that it is meaning full? How do you determine causation, or not?
Secondly, how to properly random sample a set. What sort of biases can be introduced? How can they be recognized? Documented? Eliminated (or reduced)?
Where to get your numbers to play with? Everyone has already said sports (Baseball has more stats than any other sport - and they are readily available. Watch the movie Moneyball - how stats was used to transform the management of the sport - without negatively impacting the enjoyment of watching a game). Upcoming presidential election - people prefer this candidate over that one? What sort of biases?
Yes, going through the bazillion different distributions (both discrete and continuous) is still a requirement. As well as conceptual bits as standard deviation, and the area under the curve (why two, almost identical sets of numbers can provide different outcomes).
Here's a wacky one - (yes, it deals with guns)... Find an AirSoft automatic gun... Load it with 30 rounds of bbs' (same mfgr, weight, roundness, etc.) - set it up with it locked in position 50' from a target and and just let loose the entire clip. Make 30 of these - one for each student. Perfect example of a bell curve. No matter how well you controlled the experiment you still had a distribution. A lot of info you can pull from this... mean, median, mode (which will be rings!), and if there is a vector preference - you can analyse that as well (show's bias).
Regardless - make your examples fun, make them real. And make them something they can further investigate on their own.
FredInIT
Just a few thoughts:
There is a real advantage to back in the day: parametric statistics needs calculus but a lot of modern statistics
are more simulation-based so that could be stressed. Easier to understand (or at least less difficult) and usually
more accurate. Parametric statistics can wait.
Some appropriate subset of "How to Lie with Statistics" might be apropos early on and throughout the time spent.
It's practical information in life, gives a deeper understanding and is relatively fun. Care needs to be taken that this
isn't taken as "All Statistics Lie".
Consider bringing in language teachers for help. The words in statistics often have a subtle (or huge) difference
from common usage and they may be able to help with that. I had a mathematics background when I started statistics
and wasted a lot of time in early days because "variable" meant something different than what I was used to.
Also includes more advanced ideas, like Bayes' Theorem and Central Limit Theorem, but presented conceptually.
http://www.amazon.com/Cartoon-...
https://en.wikipedia.org/wiki/...
https://en.wikipedia.org/wiki/...
This is probably only for the older students, but a good one for sparking interest is the Monty Hall problem. It has a fun narrative (if you use the goat and car), and you can set up a mock game show with some students as the contestants to get them interested. You can walk through how the game works, and then debate whether it is better to change your choice or not to maximise your chance of getting the car. Once everyone has decided, you can then run a live simulation by giving each student a turn playing and calculate the probability from the results. Most people find it really shocking when they see the probabilities are so different, and it will get them thinking there might be more to this statistics thing than boring numbers.