Reading 11: Hypothesis Testing
Reading 11: Hypothesis Testing
Reading 11: Hypothesis Testing
• Not Covered
• None
Steps in Hypothesis Testing
1. State the hypotheses: null and alternative.
• For tests concerning a population mean, the sample statistic will be the sample mean.
• For tests concerning a population variance, the sample statistic will be the sample variance.
• For tests concerning two population means, the sample statistic will be the difference of the
sample means.
• For tests concerning two population variances, the sample statistic will be the ratio of the
sample variances.
Selecting the Test Statistic—II
The probability distribution for the test statistic depends on the nature of the test statistic:
• For tests concerning a population mean, the sample statistic will follow a z-distribution or
a t-distribution.
• For tests concerning a population variance, the sample statistic will follow a chi-square
(X2) distribution.
• For tests concerning two population means, the sample statistic will follow a z-
distribution or a t-distribution.
• For tests concerning two population variances, the sample statistic will follow an F-
distribution.
Significance Level—Type I and II Errors
Selecting the appropriate level of significance means balancing two possible errors:
Generally, as P(Type I Error) decreases, P(Type II Error) increases, and vice versa. The only way to
decrease both probabilities is to increase n: to get more data.
• Data-mining bias
• Sample selection bias
• Survivorship bias
• Look-ahead bias
• Time-period bias
Calculate Test Statistic – I
• Testing one mean, z-test:
X − µ0 X − µ0
=z = or z
σ s
n n
χ 2
=
( n − 1) s 2
σ 02
Calculate Test Statistic—II
• Testing difference of two means, normal populations, t-test
• Population variances unknown but assumed equal:
t=
(X 1 )
− X 2 − ( µ1 − µ2 )
1/2
sp2 sp2
+
n n
1 2
where:
s 2
=
( n1 − 1) s12 + ( n2 − 1) s22
n1 + n2 − 2
p
t=
( X 1 )
− X 2 − ( µ1 − µ2 )
1/2
s12 s22
n +n
1 2
where:
2
s12 s22
n +n
df = 1 2
2 2
s12 s22
n n
1
+
2
n1 n2
Calculate Test Statistic—IV
• Testing mean difference of two dependent normal populations (paired comparison
test), t-test
d − µd 0
t=
sd
n
• Testing equality of variances of two normal populations, F-test
s12
F= 2
s2
• df1 = n1 – 1, df2 = n2 – 1
• df = n – 2
Decision Rule: One-Tail Test—I
• Transaction costs
• Taxes
• Risk
• Changing conditions that may make the strategy less useful in the future
p-Value
The p-value is the smallest level of significance at which the null hypothesis can be rejected.
X − µ0
z=
s
n
Step 3: Specify the level of significance
A confidence level of 95% has been given to us, so the level of significance is 100% – 95%
= 5%.
Hypothesis Testing Example—III
Step 4: State the decision rule
The critical z-value is 1.645. The decision rule is that we will fail to reject H0 if the test
statistic is less than or equal to 1.645; we will reject H0 in favor of Ha if the test statistic is
greater than 1.645.
X − µ0 1.23% − 1.0%
=z = = 1.697
s 1.05%
n 60
Hypothesis Testing Example —IV
Step 6: Make the statistical decision regarding the hypotheses
As the value of the test statistic, 1.697, is greater than the critical z-value, 1.645, we
reject H0 in favor of Ha and conclude that the mean monthly return on the fund is greater
than 1%.
The critical F-value at the 90% confidence level (10% level of significance, 5% in the upper tail) with
40 df in the numerator and 60 df in the denominator is 1.59. As 1.6767 > 1.59, reject H0 at the 10%
level of significance. The critical F-value at the 95% confidence level (5% level of significance, 2.5%
in the upper tail) is 1.74; as 1.6767 < 1.74, do not reject H0 at the 5% level of significance.
Practice Question
In a test to determine whether the average large-cap stock’s Sharpe ratio is equal to 0.60, a sample
of 40 Sharpe ratios for large-cap stocks results in a p-value of 7.9%. The null hypothesis should: