If you’re writing Knitr documents, then you need to create more variables. A lot more. I have been reviewing projects, and I am finding literally dozens of instances where an author changed his or her mind while working on whatever it was. You see something like this.
## Accuracy 0.9821 ...
The resulting accuracy is 95%.
The author must have changed something in his or her design, and accuracy has improved. The R output and the supporting text do not line up.
Or another one I’ve seen often.
## Resampling: Cross-Validated (10 fold) ...
Our training control will use k-fold cross validation, where k = 5.
Obviously, the text of the document and the R output do not match! This is so easy to fix, too. Create variables and reference them in your R code and your text. For example, if we create a
cv_num_folds variable, and reference that variable appropriately, we can avoid all future issues of code/text misalignment.
cv_num_folds <- 10 control_validation <- trainControl(method = "cv", number = cv_num_folds)
In our text, we can now write…
Our training control will use k-fold cross validation, where k = `r cv_num_folds`.
Our training control will use k-fold cross validation, where k = 10.
Please, clean up those reports!