R demo | McNemar Test | How to Conduct, Visualise and Interpret

In this video, we'll learn how and why we need a McNemar test and what happens if we wrongly use Chi-Square test instead
If you only want the code (or want to support me), consider join the channel (join button below any of the videos), because I provide the code upon members requests.
Enjoy! 🥳

[@findthetruth3021]

Your videos are always the best, easiest, useful and informative. Please make more videos like that. Could please make flexdashboard as easy as this one too or any packages can make it easier. Thanks in advance😁👏👏👏 Keep up the great work.

[@yuzaR-Data-Science]

Thanks, Halkawt! Will do! I already startet a series - package reviews, so I have a few. But wanna expand when time allows. Cheers

[@buraktiras93]

Another brilliant contribution! Thank you!
So, do you think can this test be used to evaluate a training’s before and after effects on a group of people, let’s say? What do you think?

[@yuzaR-Data-Science]

If you have categorical "before and after" data, then yes, absolutely.

[@Jonathan-zd6ns]

Awesome video! Can I ask if we have more than 2 categories (like 1-5 ordinal likert scales before and after), can we still use McNmar test to test porportional difference for paired? Thank you.
P.S. I really enjoyed your videos.

[@yuzaR-Data-Science]

Yes you can! If you wanna compare the proportions of particular likert scales, like 1s before and after vs 2s before and after you can do a general Cochran test (yuzar-blog.netlify.app/posts/2022-03-04-cochran/) and then do post-hoc pairwise McNemar tests.

[@yuzaR-Data-Science]

And thanks for a nice feedback! :)

[@odaimaihoub1222]

Great video
Thank you sir!

[@yuzaR-Data-Science]

My pleasure! Glad you liked it!

[@halkkoi1538]

This is so interesting. But, I have a data which includes pre and post test marks in number for almost 4000 people. For instance pre 10 but post 20 or the same marks or less. So I want to make something like that. Is there anyway to do so? Please help me.

[@yuzaR-Data-Science]

You'd use a mixed-effects-model. Which one and with what data is highly dependent on your hypothesis and data. I guess you contact the local statisticial, so, somebody can sit down with you, understand what your hypotheses are and consult on the type of the model. Hope that helps. Cheers