Скачать с ютуб The Suprisingly Simple Alternative to NHST: Model Comparisons в хорошем качестве

The Suprisingly Simple Alternative to NHST: Model Comparisons

Learning objectives #1. Model comparison vs. NHST approach #2. Difference between a nested and non-nested model #3. Complexity vs. fit tradeoff #4. Two reasons p-values are okay in model comparisons #5. Metrics we use to compare models #6. Know which metrics we can use for non-nested models link to NHST:    • Probability, Part 7: What is Null Hyp...   p-hacking video:    • Probability 8: What is wrong with NHS...   Technical sidenote: R square tells us the proportion of variance explained, so a 1% change in R squared means we have explained an additional 1% of the variance. But, if we equate proportion of variance explained with predictive accuracy (which actually isn’t unreasonable), I suppose you could argue a 1% change in variance explain equates to a 1% improvement in accuracy (at least relative to a model with no predictors). How’s that sound? I don’t know. I doubt anyone reads these technical side notes anyways.