I deeply respect your drive to create these lessons on statistics despite humble channel views. The videos have helped me a lot in my own academic pursuits, and continue to deepen my passion and interest for statistics. Please keep doing what you do, it’s really important!

Please, we definitely want that you make the video about how to figure out if we are hypothesising an interaction.

I don't know why your channel doesn't have more views, amazing content!

@QuantPsych

Жыл бұрын

Me neither. Must be a global conspiracy to silence those who speak truth, orchestrated by NHST advocates and the Illuminati.

@mehdiberber8395

Жыл бұрын

The most underrated Channel about stats !!!

All your content has helped me so much. please please PLEASE make a series on forecasting if it interests you!

After taking an intro stats course with R I realized I found statistics interesting and thought I wouldn’t mind doing this for the rest of my life. So glad to have found your channel as I keep learning on my own after graduating, I didn’t major in statistics. I studied biology.

I remember learning this approach in uni, and then later being asked to use anova and such tests in the lab. I was so confused. I still find model comparisons way more clear :')

fantastic video!! and very good news about your book!!

Yes, please do an interaction video as well! Consider doing it in SPSS 😅 would love to see an HLM in SPSS with output explanations

You, my man, are awesome 👊. Appreciate your videos and enthusiasm, keep up the great work!

Great video, subscribed!

Thank you very much for the great video! I have been following your content for a moment and I have been really happy to find so many practical strategies how to approach and improve the analysis of my data. By improving my understanding, you have rekindled my interest in the foundation of statistics. So thank you very much for your efforts, we appreciate them!!! And a video about interaction hypothesizing would be great!

it is the first time I find a performer and a statistician. THANK YOU 🤩 I am fascinated

Wow! Great video!! Thanks!

your lessons are fantastic, thank you so, so much!!

my mornings start always so good after I see your videos... this one just made my day AND week. Thanks for my new mantra "If you are a SPSS user, I forgive you!" 😂. In a more serious note: at minute 13:12 you define the interaction to evaluate conflict and added the interaction of climate and scarcity. Since we know both scarcity and conflict are related, could you have chosen as well to measure scarcity with the effect of climate and conflict?

@QuantPsych

Жыл бұрын

I'm not sure I understand your question...

Please do a video on interaction effects in nested models approach!!!

@QuantPsych

Жыл бұрын

kzread.info/dash/bejne/ZXd3zcd_g8uoY7g.html

Amazing!

We have domain knowledge that should dissuade us from using a constant/bias term for the drag force. Not that we couldn't build such a model, and not that such a model would necessarily be useless, but because we already have better models. The drag force on an object is not going to be best modelled with a constant term because it depends on the density of the fluid, the cross-sectional area of the object (which can be dynamic if the object rotates and isn't a ball, or deforms), and the velocity. See this wiki entry on the drag equation for more: en.wikipedia.org/wiki/Drag_equation Assuming the fluid density, velocity, and cross-sectional area are known for a collection of objects that you measured and dropped from some height, you could treat the drag coefficients as random effects for each type of object you dropped.

@galenseilis5971

Жыл бұрын

Assuming the falling object is close to its terminal velocity, and that certain variables (see wiki link) are constant, then net force could be modelled with drag force being approximately constant. en.wikipedia.org/wiki/Terminal_velocity

@QuantPsych

Жыл бұрын

Thanks for the clarification! (I'm clearly not a physicist). Also, I made a video about our nonparametric discussion from a year ago. Here's my (unlisted) video: kzread.info/dash/bejne/o5mZmMRwps-Yf5M.html I'm planning to publish it in the next few weeks and wanted to give you a chance to view it before I do.

@galenseilis5971

Жыл бұрын

@@QuantPsych Thanks for taking the time to make a response video. It is fair and gracious.

I can see why b_3*climate*scarcity is a term describing how climate affects the relationship between conflict and scarcity. To me, we can see b_1*scarcity as the term that describes the relationship, and then climate affects this since we can rewrite the terms as b_1*scarcity + b_3*climate*scarcity = (b_1+b_3*climate)*scarcity = b_1'*scarcity. But why do we introduce the term b_2*climate?

I have this book (Judd, McClelland, Ryan) using a model comparison approach but never got time to turn its pages. Thanks for the easy explanation.

@QuantPsych

Жыл бұрын

I've been reading through that one lately :)

Not to nit-pick the physics but at 10:37 , I think you're using the word 'weight' when you mean 'mass'.

@QuantPsych

Жыл бұрын

I did indeed :)

I have watched perhaps all your videos (especially on linear mixed models) multiple times since I determined that the study I'm working on would require that model. However, I have a question on determining the reduced and the full model. Shouldn't your baseline/reduced model be the one that includes all the variables in the research question? I thought that the longer model should be one that include variables that you may not yet have clear evidence that they play a role or not and then you can use model comparisons to answer that question. Secondly, if the reduced model is as you explained (barebones model after removing the variables in your hypothesis/research question), how do you then explain the result if it turns out that the reduced model (lacking variables in your research question) is the best? Secondly, a question on the second example: "What is the effect of low carb diet on weight loss once we control for calories consumed?" Since both are fixed effects, does it matter which one comes first in the statement (e.g What is the effect of calories consumed on weight loss once we control for low carb diet?) With the way the reduced model is stated, it seems there is an inherent assumption that "calories consumed" has an effect on weight loss and that we are just trying to see whether low carb diet modifies that effect. I would appreciate if you could answer these questions. Thank you very much for the excellent work you do here on KZread.

Great explanation, thanks! In some ecology studies, we have a bunch of variables and want to find out which best explain the variance in the data (including lots of interactions) - The research question therefore cannot be specific. What would you do in that case? I've sometimes reported the results from the full model and sometimes tried anova() and similar to find the best subset model, but really I have no idea what's appropriate. (Especially in cases where model selection says we can drop a variable but according to the full model results, that variable has p-value of ca. 0.01. I'm not a statistician and lack the confidence to ignore p-values and usually have no idea what the other model results mean if it's not a simple model, which it never is because ecological data is usually complex.) As an ecology student, I've probably spent more time googling statistics than reading ecology so all help is greatly appreciated. :)

@QuantPsych

Жыл бұрын

It sounds like you're in data mining territory. I have several videos on that. The idea would be to use data mining to find a plausible model, then build that model the traditional way (then replicate on an independent sample).

Great video! Some “reviewer2” suggested include in the model comparison a “null (only intercept) model” in my model comparison… what do you think about that suggestion, is this reviewer combining the old school and model comparison? (Full vs reduced model vs no relation at all “test” off null hypothesis?)

@QuantPsych

2 ай бұрын

Maybe? At least they're using model comparison language.

does the GLM have different assumptions depending on the types of variables, like there are with t-tests, ANOVA, regression etc? A step-by-step video on how to check if assumptions are met for GML would help. Thank you so much - videos are very enlightening

@QuantPsych

Жыл бұрын

Same assumptions (although the linearity assumption is always met with categorical predictors). I do have a few videos on checking assumptions. You can search my channel for that.

Omg are you actually making a textbook?!?! Please!!

@QuantPsych

Жыл бұрын

Yup! you can find it here: quantpsych.net/stats_modeling/

@seejendo3290

Жыл бұрын

And yes - please make that video - “how to figure out if you’re hypothesizing an interaction based on the language you use”

@writtenlike

Жыл бұрын

@@QuantPsych I just read the introduction and I love it! It's so striking to see that indeed, stat teachers/profs tend to teach stuff (correlation, t-test, Chi², ANOVA, multiple regression) as distinct techniques, while (apparently) it would make more sense to first see what unites all these approaches. I should definitely try to read on.

@swinginkeke

Жыл бұрын

@@QuantPsych This. Is. AMAZING. I love pen/paper, and would love to either buy a copy or print it out and bind it myself. Is the former an option? If not, any advice on compiling all the chapters from the link above and doing it myself? Also: I'm a biostatistician for a children's hospital, and I rely daily on your content to help me better do my job. Your content is phenomenal and it's beyond evident you're a great teacher. What I'm trying to say is: if you ever decide to open up some sort of QuantPsych forum/VIP membership/mentorship cohort, you can take a fistful of my money.

@pavloszournatzidis

Жыл бұрын

@@QuantPsych Your videos are so inspiring ...! Is this textbook, a good point to start for a naive stats learner who is only familiar with some basic knowledge? I am asking here for where to focus if we want to transit to GML Thanks!!

😂 when you are an SPSS user - I forgive you - but please get some help!!!! 😂 That's brilliant!

The foundation of all statistics and probability is measure theory 😅

@QuantPsych

Жыл бұрын

Let's say it has two foundations :)

@swavekbu4959

Жыл бұрын

Historically, statistical ideas existed long before measure theory came along, so it may be that the logical foundation of statistics is measure theory, but the philosophical and psychological foundations are much deeper. Statistical "concepts" go back to early civilizations. Measure theory is a logical mathematical base for statistics and probability, but true "foundations" of statistics I would argue are not found in measure theory, but rather in historical and philosophical analyses.