Using Flexplot for Mixed Models

[@maxduvall9055]

1:26 "BAM! You haven't seen this", he says while covering up the estimates we cannot see... Good one, Dr. Fife, you got us.

[@TomasMathews]

Hi Dustin, trying to use the estimates function in flexplot and it does not seem to be running unsure why this is. Any help?

[@allisonholevoet1328]

Wow. Thank you for this. While the rest of the team is out eating dinner, I'm in bed early watching your videos instead of Netflix. :) Keep up the great content.

[@QuantPsych]

I'm better than netflix!?!?!?! That's high praise :)

[@katarinawilloch8924]

Hey, great video and explanation! I was wondering if there was any way to make the main black line any smaller and the other lines appear larger/stronger?

[@高谷歌-s9s]

Hi, I'm dong a similar analysis. My model is: model

[@QuantPsych]

Please report on github.com/dustinfife/flexplot

[@AlceaRose]

I have the same problem; did you manage to fix it somehow ?

[@amberjones6306]

@@QuantPsych I am still having this problem is there any update?

[@amberjones6306]

model=lmer(darkness~ anger+ emotional_bonds + + (anger + emotional_bonds|jedi_id), + data=d) Warning message: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model failed to converge with max|grad| = 0.0248317 (tol = 0.002, component 1) > estimates(model) Error in str2lang(x) : :1:15: unexpected ')' 1: darkness~1+(1|)

[@TheBjjninja]

I love flexplot! Nice, Dr. Dustin.

[@Rockmonanov]

Not sure if this means anything to you: "Error in model$model : $ operator not defined for this S4 class", I just keep getting that anytime I use compare.fits for anything, everything else works though.

[@vornamenachname7650]

Hi I also had the problem. I think I found the reason: you cannot library/require lmerTest when using flexplot or it will interfere somehow

[@Rockmonanov]

@@vornamenachname7650 great! I thought it might have something to do with one package interfering with another. Thanks!

[@daitoc08]

Thanks very much!

[@Supertajt]

Thank you so much for your videos! Can you do a video where you explain how you would report an LMM in a table for publication?

[@QuantPsych]

Good idea!

[@brazilfootball]

Does Flexplot not work well with models built from the glmmTMB package? I get this error: Error in if (n1 < nrow(data) | n2 < nrow(data)) { : argument is of length zero

[@QuantPsych]

Probably not. I wish flexplot automatically worked with all types of models, but unfortunately, there's too much flexibility in how others can store model information. You can add it as a feature request on github: github.com/dustinfife/flexplot

[@jesteban__2612]

Hey Dustin, I'm from Colombia and I'm trying to work with agricultural data using generalized linear models. I have two questions: Is it possible to use "presence-absence" data as a response variable? If it's possible, do I need to change the link function? Thanks a lot for your videos

[@QuantPsych]

So you have a binary outcome? If so, you'll do a logistic regression (glm(y~x...., data=d, family=binomial)).

[@jessewallace-webb4475]

Thank you for creating this amazing feature! Is it going to be available in jamovi at some point too?

[@hm.91]

Best mixed models video everrrr

[@QuantPsych]

Glad you think so!

[@ryanoleary2928]

Thank you for all your videos! I was wondering, could you cover how to do linear contrasts within a mixed model?

[@evanvaldes610]

What would you recommend If I am running into convergence issues with a specific model, and I reduce the model to the point where it does converge, but model comparisons show that the more complex albeit convergence issues model is the better explaining/fitting model. How do I go about understanding that inconsistency/is it common? This is the warning: Warning: Model failed to converge with max|grad| = 0.00348609 (tol = 0.002, component 1)

[@TheBjjninja]

I'd probably use some statistical tests on the residuals to make a decision and weigh what the purpose of the model is used for also. If the residuals look better on a simpler model and I'm using the model for inference purpose and the AIC metrics look only very slightly better on a more complicated model, then I'd probably use the simpler model. If you are the using the model to make predictions, then of course use the complicated model.

[@MegaSesamStrasse]

Is there also a way to plot the residuals of the random effects with flexplot? As far as I know, these should be also normaly distributed.

[@QuantPsych]

Yes. visualize(model, plot="residuals")

[@YH-tx5qz]

Would you do mean-centering before running mixed model analysis?

[@QuantPsych]

It's often a good idea. It depends on how you want to interpret the parameters.

[@reddotmusic6251]

Thanks for these videos, they really help me a lot! I have one question (it's not the main topic of this video but I've not seen a good explanation for it also in other videos): why does it make sense to model sth as a fixed AND random effect ? Would appreciate an answer:). Keep up the great work, big fan!

[@ryanoleary2928]

This is great. Does flex plot work for bayesian models?

[@QuantPsych]

It might, but hasn't been fully tested.

[@ghaiszriki7912]

Hi (: Thaks a million man, I am almost no more scared by mixed models

[@gonzalodiazamor5494]

I am not able to download and load the library of flexplots :S If someone has had the same issue please contact me. Thanks

[@QuantPsych]

You can post on issue on github (github.com/dustinfife/flexplot). Be sure to state what kind of error message you're receiving.

[@gonzalodiazamor5494]

@@QuantPsych Thank you very much! Now is working, it was my fault sorry! btw, great work!! really cool and interesting. Thank you very much!!

[@dagmaraskova2458]

I luv u :D