STATS 13 Study Guide - Final Guide: Publication Bias, Standard Deviation, Type I And Type Ii Errors
Document Summary
Publication bias: researcher publishes findings that support their ideas but not ones that do not support their ideas. This makes the publication more statistically significant that it actually is. Polls were biased (bradley effect, likely voter model was wrong, question wording) Polls were not independent of each other (herding - pollsters only released polls that agreed with what other. Correlation (r): measures the strength and direction of a linear association between two quantitative variables. Has no units and no distinction between explanatory and response variables. 0 means no linear association; 1 or -1 means scatterplot is exactly a straight line. Rho ( ) = correlation of the population. N = people in population; = average of variable; = sd of variable pollsters have released) Given sample size n, correlation (r): greatly change the correlation minimizes the sum of the squares of the residuals. Sd of residuals = 1 2 .
