Final Exam Null Hypotheses Flashcards
μ − a = 0, where a=reference value
One sample t test
μ1 − μ2 = 0
Student’s two sample t test
μd = 0 (parametric mean difference = 0)
Paired t test
Ho: µ1 = µ2 = µ3 = µ4
ANOVA
H0: Pr{(y-a)<0} = 0.5, where a=reference value
Wilcoxon Signed-Rank test (1 sample)
H0: p{(y1 – y2) < 0} = 0.5 OR, less precisely H0: Theta d = 0 (nonparametric median difference = 0)
Wilcoxon Signed-Rank test (2 samples)
H0: p{y1 < y2} = 0.5 , or p{+}=0.5, or p{-}=0.5
Sign test
H0: p{y1 < y2} = 0.5 OR, less precisely H0: Theta 1 = Theta 2
Mann-Whitney Rank-Sum test
H0 : pi 1 − pi 2 = 0 0 < P < 1 (max two groups)
Z test of proportions
H0 : RR = 1
Relative risk
Ho = observed number of frequencies in rows & columns of contingency table are independent ; mathematically, H0: χ2=0
Chi-Square
H0 : OR = 1
Odds ratio
H0 : Pi 1 − Pi 2 = 0 or H0 : OR = 1
McNemar’s
Ho: Rho = 0
o 0 < r < 1 (positive) or -1 < r < 0 (negative)
Correlation
Ho: rho = 0
o R2 is a measure of how well the variability in y is explained by varying x
o Don’t make extrapolations beyond data set
o N.B. When you’re looking at the significance of two Hill numbers or something like that, you’re actually using a t test, so there may be a null hypothesis associated with that.
Regression
