\"Cheat Sheet\" - COMPARISON METHODS Observational study correlation Experimental correlation double blind treatment\/control groups random

"Cheat Sheet" - COMPARISON METHODS Observational study...

• Notes
• 2

This preview shows page 1 - 2 out of 2 pages.

COMPARISON METHODS Observational study: correlation Experimental: correlation – double blind, treatment/control groups, random Correlation NOT = Causation MEAN, MEDIAN, MODE Average > Median: long right tail Average < Median: long left tail Average = Median: symmetric, normal distribution STANDARD DEVIATION Data: 59, 62, 65, 68, 71 Mean: 65 Deviations: -6, -3, 0, 3, 6 Variance: (36+9+0+9+36) / 5 = 18 Square Root: sqrt (18) = 4.24 They are never negative EMPIRICAL RULE STANDARDIZATION Standard x = (x-mean)/SD PERCENTILES To find 90 th percentile: Go to 90, call it z, mark other side too 100%-20% (10 from each side) = 80% Look up 80% on z-table: z= 1.3 X percentile = z * SD + mean CORRELATION COEFFICIENT -1 < r < 1 : closer to -1 or 1 shows high correlation (+ is positive, - is negative) r = 0 means no correlation not affected by multiplying, adding, subtracting a constant REGRESSION EQUATION To calculate y given x y = y mean + r (SD y / SD x)(x – x mean ) RMSE To find x given y (in vertical strip) RMSE = SD y *sqrt (1 – r 2 ) Baseline method : use SD of y SD & REGRESSION LINES SD: tends more toward extremes