In previous posts, I described what my group presented at the 2016 Statistical Society of Canada Annual Meeting’s student case study poster competition. In a previous post, I discussed the group’s work on selecting a prediction model and the final conclusion, which turned out to be incorrect due to a mistake in our R codes. In this post, I will described the mistake as well as other considerations.
Continue reading “Predicting influenza outbreaks with Google Flu Trends: 4. Correction and comments”
As part of the 2016 Statistical Society of Canada Annual Meeting’s student case study poster competition, my group looked at the strength and timing of association between GFT estimates and reported influenza case counts as described in this post. Then, we finally built and compared multiple prediction models to predict the peak in the annual number of positive influenza tests.
Continue reading “Predicting influenza outbreaks with Google Flu Trends: 3. Prediction”
I wrote an R script to illustrate the leave-one-out cross-validation – LOOCV, and other prediction error estimation methods for bivariate data classifications for an inclass presentation during my masters program. I have also organized the demonstration into repeatable functions, which are available on my Github page. Below is an animated demonstration of the LOOCV method for LDA and KNN models for a simulated data set using the functions.
Continue reading “Leave-one-out Cross-validation”