Hypothesis testing

By:

Aya Mohamed Rabie

Lecturer at Public Health & Community Medicine
Department
Faculty of Medicine, Tanta University
Learning objectives:

1. Understand the meaning of inferential statistics.

2. Differentiate between null & alternative hypothesis.
3. Understand the concept of p value.
Definition:

Inferential Statistics are statistics that are used to infer or predict population parameters from
sample measures.

Biological phenomena are inherently variable and in this age of “evidence-based medicine” an
understanding of such variation through statistical approaches is essential not only for the
medical researcher who intends to draw inferences from his sample, but also for the practicing
clinician, and the medical teacher whose responsibility is to critically appraise the presented
inferences before accepting them into practice or the curriculum. Development of new drugs,
devices, and techniques is heavily dependent, nowadays, upon statistical analyses to prove their
effectiveness.
Applications of Inferential statistics:

I-Calculation of confidence intervals.

II-Testing hypothesis: by the suitable tests of significance.

I-Confidence intervals (CI):
 Usually the population mean and standard deviation, μ and σ (which are
called the population parameters) are unknown; all that is known is the
sample mean (X bar) and standard deviation (S); these are called the
sample statistics.
 Inferential statistics therefore involve using a statistic to estimate a
parameter by using confidence intervals.
Definition of CI:
 It is the interval of numbers that gives an estimated range of values
(around sample statistics) which is likely to include an unknown
population parameter.
 Calculation of Confidence intervals for a sample mean at 95% or at
99% confidence levels:
1-1st step is to calculate the standard error of the mean:

Standard error (SE) = SD / √ N

SD=standard deviation for data, and N = sample size

2-The 2nd step is to calculate this equation:

A-The CI at 95% confidence level = sample mean ± 1.96 SE

So, the confidence interval ranges from: Mean – 1.96 SE to mean + 1.96 SE

B-The CI at 99% confidence level = mean +2.58 SE .

So, it ranges from: Mean - 2.58 SE to mean + 2.58 SE

What is the value of 1.96 or 2.58 ? it is the value of Z in the Z normal
distribution table, at 95% or 99% confidence levels.

How to interpret the confidence interval (For example 95% Cl)

• The easy is to say that: -There is a 95% chance or 0.95 probability of

containing the true value of population parameter within this 95% CI.
II-Testing hypothesis: by the suitable tests of significance.

-Hypothesis testing is the main role of inferential statistics. It is performed through

using tests of statistical significance.

-Function (rationale) of significance tests: A significance test uses data from a sample to
show the likelihood that a hypothesis about a population is true. There are always two
mutually exclusive hypothesis, so that when the tested hypothesis is not true, then the
opposite hypothesis must be true. A measure of the evidence for or against the
hypothesis is provided by a p value.
Basic concepts of testing a hypothesis:
 The two-competing hypothesis: null hypothesis and alternative hypothesis
 The p value
 The level of significance (5% and 1% levels)
 The meaning of the statement ―there is a statistical significance difference.
 Type I and type II errors
 Power of the test
The basic concepts of testing hypothesis will be interpreted through the six steps for testing a hypothesis.

Six Steps Process of testing Hypothesis:

1-State the scientific hypothesis.

2-Formulate the statistical hypotheses.

3-Specify the decision rule to evaluate the null hypothesis.

a. Specify the test statistic, and distribution of the test statistic.
b. Determine the critical-region.
c. p value.

4-Analyze the data.

5-Evaluate the statistics.

6-Interpret the statistics.

Step 1: State the Scientific Hypothesis:

The scientific hypothesis, also known as the research hypothesis, is a testable statement
of findings that are hoped for, by the investigator. It is developed from published studies
or clinical observations.

{It should contain the target population & the relationship between the outcome/exposure
variables}.

Examples:

Early detection of prostate cancer will increase the proportion of men with prostate
cancer who survive 5 years.
Step 2: Formulate the Statistical Hypothesis

A statistical hypothesis is a statement about one or more population parameters.

Hypothesis testing includes two contradictory statistical hypotheses: the null and

alternative hypothesis.

H0: A null hypothesis usually assumes no association, no difference or no relationship

between variables.

H1: An alternative hypothesis opposes the null hypothesis, often assuming an

association, a difference or a relationship between variables.
Step 3: Specify the Decision Rule Used to Evaluate the Null Hypothesis

To specify the decision rule, identify the (a) test statistic and the distribution of the test
statistic, (b) critical region and (c) p value.

a. Identify the test statistic and the distribution of the test statistic: The significant test
statistic is computed from sample data and used to determine whether the null hypothesis
should be rejected or retained. The final value of the computed test is called the calculated
test value which is compared with a standard test value. For example, if we do a student t
test, we will compare the computed test value from the sample data, with the standard test
value in the Z distribution or the t distribution tables, at the specified p value.
b. The critical region:
 It is also referred to as the region of rejection {α}, is an area in the tail or tails of the
distribution of the test statistic.
 If the value of the computed test of significance lies in this region, this means that the
sample data is deviated from what is expected under the null hypothesis. So we must
reject the H0 and accept the H1; and the results is considered statistically significant.
 Otherwise If the results fall in the region of retention [area under the curve beyond the
tails], we retain (accept) the H0, and the results is considered statistically not
significant.
c. p value:

 A p-value is a probability value calculated when you want to evaluate two competing
hypotheses. It is the area that falls in the tail or tails of a distribution beyond the value
of the test statistic.
 Based on the normal distribution, there are standard P values, linked to the chosen
confidence level and level of significance. In any research, if the investigator choose to
adopt a confidence level of 95% and a level of significance of 5%, then the adopted
equivalent P value will be 0.05 probability, taken for the decision to reject a true null
hypothesis. Also, , if the investigator choose to adopt a confidence level of 99%, and
level of significance of 1%, then the adopted equivalent P value will be 0.01 probability.
Step 4: Analyze the data:

We analyze data either qualitative or quantitative, then compute the suitable

statistical tests so that we may determine whether the data support the null or
the alternative hypotheses.
Step 5: Evaluate the statistics:

The statistical results (step 4) are evaluated in light of the decision rules (step 3). The
investigator rejects or retains the null hypothesis. If the null hypothesis is rejected,
the investigator declares statistical significance. If the null hypothesis is retained, the
researcher concludes that no significant difference exists between treatments, drugs,
or procedures.
Step 6: Interpret the Statistics

Interpretation of significance in test results.

 If the computed value of test statistic falls in the region of rejection.(ex., P< 0.05
or P ˂0.01); then reject the H0 & accept the H1 and the results are statistically
significant .

 If the computed value of test Statistic falls in the region of retention.(ex., P˃ 0.05
or P ˃0.01); then retain(accept) the H0 & reject the H1 and the results are
statistically not significant
Other important basic concepts of Hypothesis testing:

 Types of errors:

In hypothesis testing, two errors are possible. A true hypothesis can be rejected or a false
hypothesis can be retained. These are referred to as a Type I and Type II error, respectively.

1-Type I error: this error occurs when a true null hypothesis is rejected; the probability of
such an error is noted by {α}. Type I error is similar to a false positive outcome, in which a
test result incorrectly signals the presence of disease.

2-Type II error: this error occurs when a false null hypothesis is retained. The probability of
such an error is noted by {β}. A Type II error is similar to a false negative outcome, in which
a test result incorrectly signals the absence of disease.
 Power of the test:

▪The power of a statistical test is the probability that the test correctly rejects the null
hypothesis (H0) when the alternative hypoth. (H1) is true.

▪Statistical power is the likelihood that a study will detect an effect, if the effect actually
exists. It is the ability of the test to avoid type II error.

▪The power of the test is the sensitivity of the test which is inversely related to beta Error
Type II error, so the Power increase when decrease.

▪Power of the test = 1 - Type II error

▪The best test of significance maximizes the probability of detection (this means having
high power).