See how to use bootstrap resampling to estimate model parameters using #TidyTuesday data on beer production, with code available here: juliasilge.com...
Nice presentation. Can I request if you could please show Machine learning approaches for predictive modelling using clinical covariates to predict diseases such as diabetes and cardiac arrest. Thanks
This is fantastic! Can you explain which specific violations of the assumptions of linear models would lead us to use bootstrap resampling for better estimates?
thank you, julia! question: what if I just want to unnest all splits instead of building a model? i tried unnest(map(splits, as.data.frame)) with no success
Hi Julia! Great post! I was wondering, if I did the same process, but with decision trees instead of a linear model, would that be bagging? If so, is there a way to introduce the tidy model recipe process in the model column below or should I use the the package for fitting directly? (rpart for example). Thanks! beer_models % mutate( model = map(splits, ~ lm(sugar_and_syrups ~ 0 + malt_and_malt_products, data = .)), coef_info = map(model, tidy) )
If you combine them into an aggregated prediction, yes! I like this explanation: bradleyboehmke.github.io/HOML/bagging.html The rsample package FYI was created to be super flexible for many purposes and thus isn't the most performant (if you actually wanted to use this for bagging). You might want to look into the strapgod package: github.com/DavisVaughan/strapgod
Hi Julia, thank you for your video. I am trying to apply this to my dataset and an error shows up. The bootstrapping bit worked just fine. In the next step, when applying my model (a function called anova_test from the package rstatix) to the bootstrapping: dataA.boot %>% mutate(model=map(splits,~anova_test( data=.,dv=fitA,wid=subject,between=c(exp), within=c(Oddball,PvN))), coef_info=map(model,tidy)) I get the error: Error: Problem with `mutate()` input `model`. x Can't subset columns that don't exist. x Column `fitA` doesn't exist. ℹ Input `model` is `map(...)`. I'd be really greatful if you could point me into how to fix this error. Thank you very much!
It's hard to debug this much code in comments like this. I suggest you create a reprex and post on RStudio Community, where it is easier to go back and forth to solve a problem like this one: rstd.io/tidymodels-community
@@JuliaSilge Thanks Julia! I posted it on R studio and you commented there as well, I now have the reprex, however, I am unable to upload my response on R studio (not too sure why). Would it be possible to send it some other way? Thank you very much & sorry for the inconvenience! This is the link to my question: community.rstudio.com/t/problem-using-bootstraps-from-tidyverse-with-anova-test-from-rstatix/87463/3
@@apy9484 If you are having trouble using reprex, it's definitely worth it in the long run to figure out how to use it. Check out this article that digs into some tips: reprex.tidyverse.org/articles/articles/magic-reprex.html
@@JuliaSilge Hi Julia, my issue was not creating the reprex but being able to respond to your comment on R studio. I have just created a new topic, including the reprex. Sorry for the inconvenience. community.rstudio.com/t/problem-using-bootstraps-from-tidyverse-with-anova-test-from-rstatix-2/88125 Thank you very much for your help!!
