Eric Weber made an impassioned post on LinkedIn yesterday: "Something that drives me crazy about statistics: Giving important concepts numbers or uninformative names. Example: Naming type 1 and 2 error. Half the time I forget which one is which and have to look it up."
I use mnemonics to help remember easy-to-forget statistical concepts. Here are a few. I hope you find them handy.
Type 1 & Type 2 Errors
Type 2 Error - failing to reject the null hypothesis: (Fail 2 reject, Type 2)
False Negative (False and negative - 2 negatives, Type 2)
It follows then:
Type 1 Error - rejecting the null hypothesis when it is True. False Positive.
Precision & Recall
Precision - think of Cookies ('c' in Precision and 'c' in Cookies)
Cookie-cutters help give cookies perfect shape. There is Precision; however, some dough is typically leftover.
Recall - think of Brownies. There is no dough leftover when you make brownies, but the shape is not perfect.
Bias & Variance
Bias - related to assumptions made by a model
Variance - related to Errors when you change the training data ('r' in Variance, 'r' in Errors)
Specificity & Sensitivity
Specificity - ability of a test to correctly identify patients with a disease, say Cancer ('c' in Specificity, 'c' in Correctly, 'c' in Cancer, Correctly Diagnosed with Cancer)
It follows then:
Sensitivity - ability of a test to correctly identify people without the disease
Overfitting & Underfitting
Overfitting - like overthinking. When you overthink, you see patterns that are not there.
Vineeta is an overthinker. V is also for Variance.
High Variance, Low Bias.
It follows then:
Underfitting- underwhelming analysis does not capture trends
Low Variance, High Bias.
Ridge & Lasso Regression
Ridge Regression - (2 R's) L2 regularization
It follows:
Lasso regression - L1 regularization
Please share your favorite mnemonics in the comments :)
Image Credit: Stocksy