Statistics: ANOVA

To simplify, a one-way ANalysis Of VAriance (ANOVA) is like a t-test, but for three or more groups.

Note that t-tests and ANOVAs are applied to continuous variables (i.e., comparing sample means for variables such as age, BMI, operative time, MAP, SpO₂, EBL, morphine equivalents, duty hours, ITE scores).

If the two groups are PAIRED observations (pre-op vs. post-op), then a “paired t-test” may be applied. If the three or more groups consist of REPEATED MEASURES (pre-op vs. intra-op vs. post-op vs. follow-up), then a “repeated measures ANOVA” may be applied.

Chi-square tests are applied to categorical/discrete variables (i.e., sex, obesity, diabetes, mortality, reintubation, ICU admission, board exam pass/fail). A chi-square test can be used in cases of just two groups, or in cases of three or more groups.

Other statistical tests to know include:

• Linear regression (given a set of predictors, what is the expected CONTINUOUS response).

• Logistic regression (given a set of predictors, what is the expected BINARY response).

• Mann-Whitney U-Test and Wilcoxon Signed Rank-Test (non-parametric cousins of the t-test and paired t-test, respectively).

• Fisher’s exact text (applied to the same scenarios as a chi-square test, but when the event rate is rare).