Notes: 31:30 = check for correlation across variables {corrr} 38:30 = recipe for PCA + graph contributors to PCA + perc variance explained by each PCA sdev + get PCA values juice() {tidymodels} 45:00 = reorder_within(), scale_y_reordered() {tidytext}
wow Julia, thank you very much, I have made these analyzes without Tidymodels and this blows my mind! Could you make a video with Multiple Correspondence Analysis? THANK YOU!
Hi Julia, great video. Would PLS make more sense in this case in order to see variability in respect to the target? Although it seems as step_pls does not exist within tidymodels
For the last part in your linear model, would you be able to include the rater's gender or any demographic variable/phenotype variable of the rater into the linear model? Could one do a separate PCA of the rater variables then include those components in a lm as well?
Hey Julia, thanks for sharing! I have a question. Aren't the variables mode and key qualitative? I got that they are INT in your data, but I can't see how they'd be quantitative. I'm looking forward to your answer!! I've been learning a lot with you!
I can see what you're saying here, yes, but the mode was already basically an indicator/dummy variable so not much different than what we'd do anyway. For key, I could definitely see your argument that it should be treated qualitatively (is that really a meaningfully continuous variable?) in which case we would want to create dummy/indicator variables with it.
