6 Comments
User's avatar
Ambika Gopalan's avatar

This is awesome! Quick question -- I'm wondering how much of your lessons incorporated acknowledging the huge assumptions anyone who uses any kind of stats makes (and the fact that you can't just "check" if they're "true" or "false"). As a practicing statistician and social science researcher I find that it's common to take a very formulaic approach to stats -- just plug in numbers into formulas, accept and reject hypotheses, and you're done! -- which is psychologically pleasing but ultimately a bad approach. Perhaps another area to evolve stats education?

Ethan Ludwin-Peery's avatar

Thank you! It's a good question. I spent the last week of class trying to teach them how parametric and non-parametric tests work at a mathematical level, so I think we covered a lot of the assumptions implicitly, though we probably didn't cover them in the way that you mean. I do think taking a look under the hood is one way to fight the formulaic approach, but at the same time it's true that they didn't have any lessons explicitly about how to check assumptions. It's something that would be good to add in a longer or more advanced course!

Julian's avatar

Wow, I'm very curious to see what the actual exam questions were like!

Could you post a pdf?

Ethan Ludwin-Peery's avatar

Sure! I have all the answer keys. Here's one as an example, see if you can get the answers without checking!

https://drive.google.com/file/d/1cmAfr0npbJ-8O_xR0QdtOXQ-wWWIZgZ3/view?usp=sharing

Julian's avatar

This was fun!

I was able to get most of the questions correct, with the exceptions of cereal_whattest

castle_whattest

formula_whattest,

because I forgot the difference between z-tests and t-tests and I've never actually studied ANOVA

I like how you used multiple choice questions to remove ambiguity

in SPSS_nullh and table_ord, my initial answer was different from the official correct answer, but it wasn't a provided option, so I wouldn't have lost points.

diff_power was tricky because I almost misread p = .48 as p = .048, because I'm much more used to seeing the latter :)

I found p_from_CI tricky because it wasn't obvious to me that a slope would have a non skewed distribution. So I might have selected can't tell for "The 10% confidence interval includes 232"

Julian's avatar

Yay, thanks, will try it!