Scribd
Upload
17 views

TP04 BasicStatistics p3

This document discusses hypothesis testing in statistics including the concepts of null and alternative hypotheses, types of errors, test statistics, and decision rules for hypothesis testing. It provides an example of testing the hypothesis that the mean risk premium on stocks is greater than or equal to zero using a t-test and defines critical values for different significance levels and hypothesis types.

Uploaded by

tommy.santos2
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
17 views

TP04 BasicStatistics p3

This document discusses hypothesis testing in statistics including the concepts of null and alternative hypotheses, types of errors, test statistics, and decision rules for hypothesis testing. It provides an example of testing the hypothesis that the mean risk premium on stocks is greater than or equal to zero using a t-test and defines critical values for different significance levels and hypothesis types.

Uploaded by

tommy.santos2
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Data Analysis in Finance

Master in Finance

Class #4
Jorge Caiado, PhD Luís Silveira Santos, PhD
CEMAPRE/ISEG, University of CEMAPRE/ISEG, University of
Lisbon Lisbon
Email: [email protected] Email: [email protected]
Web:
http://jcaiado100.wixsite.com/
jorgecaiado
II. Basic Statistics
Hypothesis Testing
Estimation and hypothesis testing constitute the two branches of statistical inference.
Estimation address the question: “What is the value of the parameter θ?”. Having
obtained a random sample of size n, we build a confidence interval for the true
parameter θ around the estimator θ. � Since the true θ is unknown or rarely known, we
raise the question: Is the estimator θ� compatible with some hypothesized value of θ,
say θ = θ∗ ?

In statistical inference language, θ = θ∗ is called the null hypothesis (H0) and is tested
against and alternative hypothesis (H1), ≠ θ∗ .

The null hypothesis and the alternative hypothesis can be simple or composite. The
hypothesis is simple if it specifies the value(s) of the parameter(s) of the distribution,
for example 𝐻𝐻0 : θ = 4. The hypothesis is composite if the value(s) of the parameter(s)
of the distribution is not specified, for example 𝐻𝐻0 : θ ≥ 4.

To test the null hypothesis, we use the sample information to obtain the so-called test
statistic.
2
II. Basic Statistics
Example: economic theory suggests that investors require a positive risk premium on
stocks. We can formulate the null and alternative hypothesis as follows:
𝐻𝐻0 : μ𝑅𝑅𝑅𝑅 ≤ 0 vs. 𝐻𝐻1 : μ𝑅𝑅𝑅𝑅 > 0
According to Dimson et al. (2011), the annual mean equity risk premium for Canada
stocks relative to bond stocks from 1900 to 2010 (111 annual observations) was 𝑥𝑥̅ =
5.3% and the standard deviation for the annual risk premium was s = 18.2%. Since
the risk premium (𝑋𝑋𝑖𝑖 ) follows a normal distribution, 𝑋𝑋𝑖𝑖 ~ 𝑁𝑁(𝜇𝜇, 𝜎𝜎 2 ), we know that the
test statistic 𝑋𝑋� is distributed as 𝑋𝑋� ~ 𝑁𝑁(𝜇𝜇, 𝜎𝜎 2 /𝑛𝑛). Standardizing the normal variable, it
follows that
𝑋𝑋� − 𝜇𝜇
𝑍𝑍 = ~ 𝑁𝑁(0,1)
𝜎𝜎/ 𝑛𝑛
or, because we do not know the population standard deviation,
𝑋𝑋� − 𝜇𝜇
𝑇𝑇 = ~ 𝑡𝑡(𝑛𝑛 − 1)
𝑆𝑆′/ 𝑛𝑛
Note that, with many observations, 𝑡𝑡 𝑛𝑛 − 1 will converge to 𝑁𝑁(0,1).

NOTE: We may use confidence intervals or test of significance approach for


hypothesis testing
3
II. Basic Statistics
In deciding to reject or not reject 𝐻𝐻0 , there are four possible outcomes:
 We reject 𝐻𝐻0 when it is false: No error or correct decision.
 We reject 𝐻𝐻0 when it is true: Type I error.
 We do not reject 𝐻𝐻0 when it is false: Type II error.
 We do not reject 𝐻𝐻0 when it is true: No error or correct decision.

We keep the probability of committing a Type I Error at a low level, such as 0.01, 0.05
or 0.1, and try to minimize the probability of having a Type II Error as much as
possible.

The probability of a Type I Error is also known as the significance level, denoted by α,
and the probability of a Type II Error is denoted as β. The probability of not
committing a Type II Error is 1 − β and is called the power of the test.

The classic approach to hypothesis testing is to fix the significance level α as 1% (α =


0.01), 5% (α = 0.05) or 10% (α = 0.1) and then try to maximize the power of the
test, that is, to minimize β.

4
II. Basic Statistics
The decision rule on whether to reject or not reject 𝐻𝐻0 can be based on:
1) the comparison of the observed (or calculated) value of the test statistic and a
critical value;
2) The p-value.

The critical value depends on the significance level, α, and on the type of the test:
• For 𝐻𝐻0 : 𝜃𝜃 = 0 vs. 𝐻𝐻1 : 𝜃𝜃 ≠ 0 (two-tailed hypothesis test), two rejection points exist
(one negative, one positive). At the 10% level, use 𝑧𝑧𝛼𝛼/2 = ±1.645; at the 5% level,
use 𝑧𝑧𝛼𝛼/2 = ±1.960; at the 1% level, use 𝑧𝑧𝛼𝛼/2 = ±2.576.
• For 𝐻𝐻0 : 𝜃𝜃 = 0 or 𝐻𝐻0 : 𝜃𝜃 ≤ 0 vs. 𝐻𝐻1 : 𝜃𝜃 > 0 (right-tailed hypothesis test), one
rejection point exists (positive only). At the 10% level, use 𝑧𝑧𝛼𝛼 = 1.282; at the 5%
level, use 𝑧𝑧𝛼𝛼 = 1.645; at the 1% level, use 𝑧𝑧𝛼𝛼 = 2.326.
• For 𝐻𝐻0 : 𝜃𝜃 = 0 or 𝐻𝐻0 : 𝜃𝜃 ≥ 0 vs. 𝐻𝐻1 : 𝜃𝜃 < 0 (left-tailed hypothesis test), one
rejection point exists (negative only). At the 10% level, use 𝑧𝑧𝛼𝛼 = −1.282; at the
5% level, use 𝑧𝑧𝛼𝛼 = −1.645; at the 1% level, use 𝑧𝑧𝛼𝛼 = −2.326.

The p-value is the smallest significance level at which 𝐻𝐻0 can be rejected. The smaller
the p-value, the stronger the evidence against 𝐻𝐻0 . For a p-value less than α, we reject
𝐻𝐻0 . For a p-value greater than α, we do not reject 𝐻𝐻0 . 5
II. Basic Statistics

Exercise: (Source: DeFusco et al., 2015)


Willco is a manufacturer in a mature cyclical industry. During the most recent
industry cycle, its net income averaged $30 million per year with a standard
deviation of $10 million (n = 6 observations). Management claims that Willco’s
performance during the most recent cycle results from new approaches and that we
can dismiss profitability expectations based on its average or normalized earnings of
$24 million per year in prior cycles.
a) With μ as the population value of mean annual net income, formulate null and
alternative hypotheses consistent with testing Willco management’s claim.
b) Assuming that Willco’s net income is at least approximately normally distributed,
identify the appropriate test statistic.
c) Identify the rejection point or points at the 0.05 level of significance for the
hypothesis tested in Part a).
d) Determine whether or not to reject the null hypothesis at the 0.05 significance
level.

You might also like