Intro Statistics Doesn't Have to be Confusing!
Вставка
- Опубліковано 7 лют 2025
- January Mixed Models Class: simplistics.ne...
February Simplistics Class: simplistics.ne...
To see a list of all classes I offer: simplistics.ne...
This video is based on this paper: osf.io/preprin...
That paper contains the references mentioned in the video.
See my textbook at: simplistics.ne...
🎯 Nailed it again!👏
It's important for me to repeat that I'm not a Statistician because what I know is what I stumbled upon, not what I was taught.
And, in trying to understand stats, I noticed that the formulas (that's what I call them as a lay person) all seemed to follow a LINEAR MODEL PATTERN (for the most part at least).
Then that recurring pattern caused me to remember: y = mx + c when I was in high school.
Then there was an "ah hah" moment which occurred (many moons later that got me upset because nobody taught me that the high school equation was actually a linear model 😢), and the "ah hah" triggered me to start looking at data in terms of dependent and independent variables.
And now, thanks to you, my hunch and self-taught approach to reviewing datasets have been confirmed!
I know I've said it before, but you're an awesome professor and statistician!👏✨
Quite brilliant actually. 👍
Thanks! I had a similar ah hah moment later than it should have happened. I realized that stats is just algebra + error :)
@@QuantPsych Honestly, you're a veritable Genius!✨✨✨ Wow!👏👏👏 Seriously on point!🎯
the old problem of generalisation producing simplification which can only occur after seeing a 'big (complicated) picture'. Our brains keep working backassward. It's why (one of) the reasons I try and read text books backwards
Thank you!
Academia is supposed to advance knowlege but it seems academia itself is resistant to adopting or implementing the new knowledge that is a product of its own normal function. And sometimes folks in academia wonder why people outside are slow at adopting their "academic inventions".
Ah, the irony.
Looking forward to reading the paper. Definitely agree on the pedagogy of stats. There's no question that it could better. Having been a TA for intro stats now, I have seen it first hand. I think "inertia" is definitely a good way to describe the situation. It's going to take a lot of effort to change, but that effort will be spread across many hands. I think we'll get there.
I'm working on some shiny dashboards for a design of experiments course with a prof at my uni. The text in the course is titled "a model comparison perspective". It's definitely looking at ANOVA through the lens of model comparisons. It's a good book, but don't try to pick it up, it'll break your back! I thought about your approach when I saw that was the text for the course. We read a good article by Wasserstein that you have probably read. That also reminded me of your vids. I guess my point is that there are definitely like minded folks out there doing work to move the needle on this.
Is that Maxwell and Delaney's book? That was my graduate stats books. Great book!
Mostly it's teaching the ability to communicate with those who went before, hence that's what's taught. Communication is hard, especially with inappropriate mental models!
@@QuantPsych it's the one with Kelley as a coauthor (We just call it MDK for short). You cite it in your paper (Designing Experiments and Analysing Data, A Model Comparison Perspective). Looking at the Jacket, yeah, same same. I have the 3rd edition.
I never had stats in my undergrad, and then in my Master's the prof told us to "forget all the previous curriculum", and jumped right into teaching us R while following a model comparison approach book by Judd & McClelland.
I used to think that I'd be at a disadvantage for not having learned stats in my Bachelor while other students did, but I was less confused than anyone else xd
Nah, you were ready to go :)
The paper referenced in this video seems to have a broken link. Could you please share the title of the paper and its authors? This would help us locate and read the paper ourselves. Thank you.
Prof, you know how I am going to answer your question. Why are we doing things the old way? Tradition? Sure. But the real problem is with communicating to people. The solution is not to "push back" but to be patient, hear them out and then ask them what they think about the LM approach. The solution is not an us-vs-them mentality, but acknowledging complexity and patience and then hopefully everyone can learn from the dialogue that happens from that.
I know that from your perspective, as someone who has been pushing for a long time, that this must sound absurd. But you may be surprised at what you find out and how much people are willing to change when you listen, repeat what they say back to them and then ask questions. All I have to ask is, have you tried this?
I have. Sometimes I present as more uncompromising than I actually am, probably because it's more concise to do say
Interesting!! That reminds me of this verse(s) "If they are told “Follow what God has revealed,” they say, “Rather we will follow what we found our fathers doing,” even though their fathers did not understand anything and were not guided."
Couldn't agree more. Also didn't realise you developed the Visual Modeling module for JASP - very cool.
Personally I find intro to stats so hard in the psychology curriculum because the majority of 1st year undergraduate students don't take psychology to learn statistics. Or even expect to have to learn statistics.
I suppose my question is (and this is largely driven by where I receive pushback on changing the curriculum): what is your response to those who (possibly rightfully) worry that by only teaching a GLM approach, much of the scientific evidence students read in both statistics courses and other subjects, becomes more difficult to interpret, as the majority of published research in psychology has not adopted this approach?
As you say in your video, even reviewers who are supposed to be highly educated peers are not on board with using GLM as substitutes for t-tests/ANOVAs. I experience the same thing when publishing using Bayesian methods, and have been asked more than once to provide a complementary frequentist analysis to aid readers interpretation of the results.
For me, this is one of the biggest hurdles in changing from the 50+ year old traditions. Along with other academics not wanting to have to also learn the methods in order to teach them...
Great question, and I think it's the only valid criticism of my method. What I do is teach students the "old names" for things. So, I'll say something like, "this is a linear model with a categorical predictor with 3 groups, which we used to call ANOVA." I also encourage students to publish their papers by writing using linear model terms, but some are more comfortable using the old names.
@@QuantPsych Thanks for the reply! Totally agree. I think it also speaks to the issues you outline of stats teaching not changing in the past 50 years. I think the 'cookbook' approach that has become the norm in psychology is a terrible way to think about statistics, particularly as (at least in my experience) most of the clients I work with are more interested in prediction and the quantification of uncertainty, rather than just differences between groups, which is much better suited to GLM approaches. I find GLM output is also much easier to describe to clients with non-psych backgrounds as well, so when I transitioned into an academic position I was surprised to see the same stuff I learned in undergrad. I've downloaded your pre-print and am working my way through the Order of the Statistical Jedi book as well, hopefully will give me some support in changing our stats curriculum.
Been preaching the truth since 2020!
Great video! Peace out
Hi, thanks for the great content, can you recommend books about this subject? Thanks.
My textbook! It's linked in the description.
"Just" need to swap the teachers. Let those that taught advanced now teach introductory stats ('cos they know..), and let the intro teachers teach advanced stats (which is just a few tweaks to the now wonderful, all encompassing, intro stats). What could go wrong? [but why, but why ;-)
Great rants. keep it up.
Love this idea :)
@@QuantPsych then swap the students... Advanced first, "Intro" second (or perhaps "Fundamentals", to make it sound harder 😁.
I like to 'read' text books from back to front (speed read; I know I won't fully understand, but can see ho the 'missing' bits fit in, i.e. next chapter)
Hello Dustin. When i try to view the book from my phone, i can see the TOC sidebar but it takes up the whole page and i cant move it to the side, so the text isn’t visible (it’s cut off). Can you please look into it.
I'm searching around the description for the link to your free textbook, but I can't find it. Which link should I use?
Or even the non-free textbook. I was hoping there'd be a link.
I just added it to the description (simplistics.net/stats_modeling/)
@@QuantPsych Thanks! 👍
@@QuantPsych Thanks!
@@QuantPsych Is there a downloadable pdf version available? I saw on Github there is a bookdown option to create one from source and I tried to do this. However, I couldn't get past not being able to install the flexplot package (which it said was unavailable for this version of R (which I also updated from R-4.0.5 to R-4.4.2)). Any pointers gratefully received!
I would like to nominate the preview image of this video for the Pulitzer Prize.
I second your nomination :)
I arrived first. I have no comment yet because i have not finished watching this video posted 2-minue-old ago
That's a real deep comment...deeper even than your second comment :)
A Poisson result if ever I saw one. Delivering half a baby every 4.5 months ;-)