Statistical Tests and Shit

Question 1

Odds (eq and notable point)

Answer 1

p of x occurring/(1 - p)

Difference b/w odds and probability gets bigger as events become more common

Question 2

RR vs. OR Equation and Event Size

Answer 2

As a and c become smaller (events less frequent), RR starts to resemble OR eq. So THE TWO APPROXIMATE EACH OTHER WHEN EVENTS ARE RARE, BUT NOT WHEN EVENTS ARE COMMON

Question 3

RR vs. OR in terms of failure to occur

Answer 3

RR is not symmetrical when measuring disease’s occurrence or failure to occur, whereas OR is

Question 4

T Test

Answer 4

Basically a comparison of two means over variance (so dependent on N), used to generate a p value and determine if statistically significant difference

Question 5

CI in T Tests

Answer 5

Significant if don’t contain null value of 0 (no difference in means)

Question 6

Condition Required for T Tests

Answer 6

2 samples have to have equal/similar variances

Question 7

Paired-Sample Mean Comparisons

Answer 7

T test where mean difference is now difference b/w two scores for same person (beginning and end)

Question 8

Contingency (2x2) Table

Answer 8

Determine null (expected) value for each cell by doing (Total row x Total column) / Total

Question 9

3 Tests to Determine Significance from Different Types of Variables

Answer 9

2 Categorical Variables - Chi Squared
2 Continuous Variables - Correlation
1 Each - T Test

Question 10

Correlation

Answer 10

Used for 2 continuous variables just to show association. Not causation. Can determine significance from null r = 0 by using DoF to calculate t

Question 11

Linear Regression

Answer 11

Study of relationships b/w single outcome variable and 1+ exposure variables. y = intercept + X*beta coefficient (slope, just like y = mx + b)

Question 12

R squared

Answer 12

Tells what percentage of the variance around the mean is explained by changes to the predictor variable we’re looking at

Question 13

Linear Equations to Adjust for Confounding

Answer 13

Addition of another independent variable to the model suggests confounding if after addition of new variable, the beta coefficient of your primary predictor variable changes significantly

Question 14

Logistic Regression (4)

Answer 14

If outcomes are binary (often the case in medicine), log transformation allows calculating probability of outcome occurring and calculate the OR. Can control for confounding

Question 15

Survival Analysis

Answer 15

Time until an event - survival time

Question 16

Kaplan Meier Plot (3)

Answer 16

Used in survival analysis - usually interpreted for median survival time and survival percentage at specific time (i.e. 6 month survival)

Question 17

Log Rank Test

Answer 17

Compares survival experience of 2 arms of survival test and generates p value determining if difference significant or not

Question 18

Censoring

Answer 18

Time to event not observed in all subjects, so lost patients are classified as “censored” which is neither alive or dead

Question 19

Cox Regression Analysis

Answer 19

Multivariate methods applied to survival data to control for confounding

Question 20

Hazard Rate and Ratio

Answer 20

Hazard rates show instantaneous risk of not surviving, and then can make as a ratio to make a RR of hazard rates