Examples for Hypothesis Testing
- A newspaper claims that over 40% of college undergrads have wild parties before they obtain their bachelor’s degree
- State Null hypothesis, State Alternative Hypothesis
- A statistician performs study and the proportion from the sample he was working with turned out to be 0.45. The probability of observing a value of 0.45 or more (p-value) in such sample is 0.20 when assuming that null hypothesis is true. Do you have strong evidence to reject the Null Hypothesis?
a) Ho: p=.40
Ha: p > 0.40
2) p0 = 0.40
phat = 0.45
p-vaue = 0.20
Since the level of significance is not mentioned it is assumed to be 0.05, so we would reject H0 if p-value < 0.05. which is not the case here. Conclusion: we FAIL TO REJECT null hypothesis concluding that there is no enough evidence to say that the proportion of undergrads is more than 0.40
- It is important for travelers that mass transit vehicles exhibit a high percentage of on-time performance. Suppose a regional train system boasts that its trains run on time over 80% of the time. To check out this claim investigator collected the sample of real data, and sample of 40 trains arrivals was observed 37 of which came on time.
- What’s population proportion in this problem
- What claim is being made about this population proportion?
- Stat null and alternative hypothesis
- Would you reject or not Ho?
Ho: p=.80
Ha: p > 0.80
phat=37/40 = .925
po = .80
z-statistic = (phat - po) / sqrt(po (1-po)/n) = (.925 - .80)/sqrt(.80(.20)/40) = .125/.0632 = 1.97
Since testing Ha: p >.80, p-value is the area in the upper tail, or to the right of z-value.
Looking at the Table, find percentile associated with z=1.97 = 2.0 97.73%, but that’s area to the left, since we need area to the right we do: 1- .9773 = 0.0227
p-value = 0.0227
By default we compare p-value with 0.05 significance level and see whether
p-value < 0.05? Yes it is. Conclusion: We reject Ho, stating that there is enough evidence to say that trains run on time more than 80% of time.
- Records from several year ago show that in a certain community 60% of the people owned their houses. An advocacy group for affordable housing claims that home ownership has declined since that time. A survey of 50 families show that 9 own their home. Is this conclusive evidence that there has been a decrease in home ownership in the community?
Ho: p=0.60
Ho: p < 0.60
phat = 9/50 = 0.18
po= 0.60
z-value = (phat - po) / sqrt(po (1-po)/n) = (0.18 – 0.60) / (sqrt(0.60*0.40/50)) =
- 0.42/ 0.0692 = - 6.069
Perventile associated with z= - 6 is not in the Table B. For all the values of Z< - 3.4 the percentiles will be less than 0.03%, . We will take 0.02% for z = - 6
Since our Alternative hypothesis has sign ‘<’ meaning that p-value is in the left (lower) tail, the p-value is 0.02% / 100 =0.0002 which is very small compared to 0.05 significance level. Conclusion: we REJECT Null hypothesis giving evidence that there has been a decrease in home ownership.
- Public Health Service examined a representative cross section of several thousand men aged 18 to 74. The systolic blood pressure of these men followed normal distribution with mean mu =129 and standard deviation, sigma = 18. Suppose we took a random sample of 44 women aged 18 to 74 and found that their average systolic blood pressure is xbar = 120 with st dev s = 21. Does this prove that, on average women have lower systolic blood pressure than men?
Ha: mu <129
H0: mu = 129
mu0 = 129
xbar = 120
s=21
n =44
z-value = (xbar – mu0)/(s/sqrt(n)) = (120 – 129) / 21/sqrt(44) = - 2.84
p-value = 0.26%/100 = 0.0026 since it is in lower tail.
p-value is less than 0.05, meaning that we REJECT hull hypothesis, concluding that there is strong evidence to say that on average women have lower systolic blood pressure.