Multiple linear regression with interaction in R
Вставка
- Опубліковано 14 гру 2021
- Need to include interaction with a categorical variable in your linear model in R? Start here.
If this vid helps you, please help me a tiny bit by mashing that 'like' button. For more #rstats joy, crush that 'subscribe' button!
You can find the script from this vid (and others) at github.com/equitable-equations/youtube.
THANK YOU. This is the clearest explanation I've found. It's not that complicated but for some reason it's so hard to find a clear explanation.
I know, right?!
Very helpful. Thanks.
great video really well explained.
Hey, thanks a lot for this video!! Made my study a bit clearer. However, I'm struggling with my data bundle, and I'd like to ask for your help.
In my case, the equivalent to bodymass would be transcripts per million (TPM: the amount of normalized transcripts of a specific gene). Flipper length would be a categorical variable, 'Exponential' and 'Stationary', and the Species would be the Treatment to which the cells have been submitted (Solid or Liquid growth). I'm struggling with the concept of the slope because my x variable is not numerical. Would you have some input about this? Thank you very much!
Thank you!
Great video we hope more about logistic regression
Yes! It's on my list, I promise. I have a few others to do first, I'm afraid.
Multiple logistic regression too, please! :D
Impressive video, thank you so much.
Q. how to add confidence interval, statistics like p-value and correlation in multiple regression? Thank for your time and collaboration.
great video
Thanks!
Great video, thank you! I'm confused though about the equation for the theory vs. in R. Why is the equation y = b0 + b1x1 +b2x2 +b3x1*x2, but we write it in R as glm(x1 ~ x2*x3...) instead of glm(x1 ~ x2 + x3 + x2*x3...)? As in: body_mass_g ~ flipper_length_mm + species + flipper_length_mm * species? Hopefully that's clear, haha.
Hi! First off, there's no need to use glm here instead of lm since there's no link function involved. Regardless, either command will include individual terms all interaction variables by default when you use *. This is generally best practice, but if you need to override it, you can do so but replacing x2*x3 with x2:x3.
@@EquitableEquations Ah, yes, sorry for confusion. I copied the glm part from my code. Thank you!
Helped with a conference paper. Thanks
Can you do multiple interactions with this method? Like if you wanted to split up the data for flipper length for male Chinstrap, female Chinstrap, male Gentoo, and female Gentoo so you get 4 lines to compare to each other.
Sure, but I've always avoided this in practice. Overfitting is a risk.
Thanks very much for the great instructional video. Much appreciated! Can you help me with a minor detail please?
I have 4 levels of a categorical variable('biostimulant') and my linear model summary output lists this variable by name then a number eg, Biostimulant1, Biostimulant2 etc.
How does one identify what the variable name is, that R has assigned 1,2,3,4?
Thanks in advance!
Hi! R goes alphabetically by default. Use summary() to see the details of the model with variables listed by name.
Here is my question. How do we know which is which in terms of the dependent and independent variables? In other words which is true? Flipper length controls body mass or body mass controls flipper length?
Would you recommend mean-centering "flipper_length_mm" first or is it not necessary? If so, can you recommend a resource for how to do it in R and how to interpret the results?
Off the top of my head I can’t think of a reason it would be helpful in this setting. But it’s a wide world full of mystery and wonder, and I only understand little corners of it.
@@EquitableEquations No problem, thank you 😁
@@EquitableEquations in a related question, if there are two scale variables, does R automatically mean-center before calculating the interaction or do we need to do this first before using the lm() function?
If you had say 5 dependant variables and a 2 level factor variables, say species, do you need to multiply every dependent by species in the lm function?
Model selection is really complicated! In the end you just have to respect the data (though all else being equal I like simpler models without lots of higher-order terms).
dang i wish u were my teacher
You look and sound like little finger from game of thrones
Ouch dude 🤣🤣🤣