Models
Statistical Significance
our data is almost always collected from a sample
If we observe something unusual in the data
The observed sample relationship may have occurred because of a real relationship in the population - statistically significant
The observed sample relationship may have resulted due to chance
H0: The Null Hypothesis:
The status quo: an established claim
presumed true unless strong evidence to the contrary is demonstrated
HA: The alternative hypothesis
a rival claim
what the person doing the test wants to prove
must be “proved” true
conclusions
One hypothesis must be true and the other must be false
alpha = P(type 1 error) = P(reject H0 | H0 true)
confidence level = 1-alpha = P(accept H0 | H0 true)
beta = P(type 2 error) = P(accept H0 | H0 false)
Power = 1 - beta
Test Statistic
the statistic that we use to decide between the two hypothesis
what values f the statistic will support the alternative hypothesis?
reject the null if test statistics are sufficient
H0: Coin is fair → P(H) = p = .5
HA: Coin is not fair → P(H) = p ≠ .5
Test statistic: sample proportion of heads = phat
Decision rule: reject H0 is phat is unusually large or is unusually small
P-values
A ,measure of the strength of the evidence in favor of HA. The smaller the p-value, the stronger the evidence in favor of HA
Sampling distribution of Phat
Phat = sample proportion
original population has Population proportion = p
Sampling distribution of Phat for samples of size n will have…
mean = p
Standard dev = sqrt(p(1-p)/n)
for “large samples",” will be approximately normal
NULL hypothesis: population mean = u
test statistic: xBar
Standardized: z = (xBar-u)/(SampleStdDev/sqrt(n))
is population proportion = p
test statistic sample proportion = pHat
Standardized: z = (pHat - p)/sqrt(p(1-p)/n)
Is a coin biased in favor of heads?
H0: coin is fair → P(H) = p = .5
HA: Coin is biased towards heads → P(H) = p > .5
p-value = area to the right of z = 1-pnorm(z)
reject H0 if p-value <alpha
accept H0 is p-value >= alpha
Is a coin fair?
H0: coin is fair → P(H) = p = .5
HA: noic is not fai → P(H) = p ≠ .5
p-value = 2(pnorm(-abs(z))) = 2(1-pnorm(abs(z)))
reject H0 is p-value < alpha
accept H0 if p-value >= alpha