Hypothesis

Testing
Ibrahim Altubasi, PT, PhD
The University of Jordan
 Hypothesis Testing
Example 1:
Research question: Does stimulation during infancy have a beneficial effect on
human development?

Background: observation of infant rats; increased handling results in eyes

opening sooner, more rapid brain maturation, faster growth, and larger body
weight.

The way to answer this research question: ???

From national health statistics: the distribution of weight for 2-year-old

children is normal with mean μ=26 pounds and SD σ=4 pounds
 Hypothesis Testing
Research hypothesis:
increased stimulation Research hypothesis is the
early in life can be hypothesis the researcher has
beneficial. Specifically, about his/her research
increased handling of question(s) or research
babies will result in variable(s). This hypothesis is
larger body weights. usually based on existing
research literature, on
observation, or through
reasoning.
 Hypothesis Testing
The way to answer the research question or to test the research hypothesis:
inferential statistics

The goal of inferential statistics is to make general statements about the

population by using sample data.

What we know?

From national health statistics: the distribution of weight for 2-year-old

children is normal with mean μ=26 pounds and SD σ=4 pounds
 Hypothesis Testing
Research process (data collection):

A random sample of 16 new born babies

Parents were trained to provide the additional daily handling to the

baby

The body weight for each of the 16 infants were measured at 2 years
of age

The mean weight of this treated sample is x̅ = 29 pounds and sample

SD is s=3.8.
 Hypothesis Testing
We want to know, through analyzing the data we collected, whether the
treated population on average has a significantly larger body weight compared
to the original population.
 Hypothesis Testing
Statistical hypothesis:

Null hypothesis (H0): μ=26

The null hypothesis

• states that the treatment has no effect

• is denoted by the symbol H0

• is always stated in terms of population

parameters.
 Hypothesis Testing
Statistical hypothesis:

Null hypothesis (H0): μ=26

Alternative hypothesis (H1) or (Ha): μ > 26

The alternative hypothesis

• states that the treatment has effect

• is denoted by the symbol (H1) or (Ha)

• is always stated in terms of population

parameters.
 Hypothesis Testing
Statistical hypothesis:

Null hypothesis (H0): μ=26

Alternative hypothesis (H1) or (Ha): μ > 26

The Null hypothesis (H0) The alternative hypothesis

states that in the population, (H1) states that there is an
there is no effect, no effect, there is a difference, or
difference, or no there is a relationship in the
relationship. population.

The Null hypothesis and the alternative hypothesis are mutually exclusive
and exhaustive. They cannot both be true.
 Hypothesis Testing
Hypothesis testing is a statistical procedure that uses data from a sample to
test or evaluate whether a statistical hypothesis (Null hypothesis) about a
population parameter is true or not.

Null hypothesis (H0): μ=26

Alternative hypothesis (H1) or(Ha): μ > 26

How likely would we get a sample mean x̅ = 29 pounds or larger from a

random sample (n=16) if the null hypothesis is true?
 Hypothesis Testing
If the null hypothesis is true, then the distribution of weight for
2-year-old (treated) children should be normal with mean μ=26
pounds and SD σ=4 pounds

The sampling distribution of the sample means with n=16 should be normally
distributed with mean μx̅ = 26 pounds and
SE σx̅ = 4/√16 = 1 pound.

Our sample mean x̅ = 29 with sample size n=16. Where is this sample mean
located in the sampling distribution of the sample means with n=16 if H 0 is
true? What is the probability that sample mean is as large as or larger than 29 if
H0 is true?
 Hypothesis Testing
Zx̅ = x̅ - μx̅ / σx̅ (Z- statistic)

In this study, z = (29-26) / 1 = 3, p(z≥3) = .0013

If the null hypothesis is true, the probability that we will get a z statistic this
large (z=3) or larger from a random sampling is .0013, which is so small that we
reject the null hypothesis. As a result, we conclude that the additional handling
of the babies increases the 2-year-old children’s body weight.
 Hypothesis Testing
In this study, z = (29-26) / 1 = 3, p(z≥3) = .0013

In other words, if the null hypothesis is true (in our example, if the treated
population mean is the same as the original population mean, which is 26
pounds), it is very unlikely that we will get a random sample of size 16 from
this population whose average weight is 29 pounds or larger.

Because it is so unlikely, we reject the null hypothesis that the treated

population mean equals to the original population mean and conclude that
the additional handling of the babies increases the 2-year-old children’s body
weight.
 Hypothesis Testing
Zx̅ = x̅ - μx̅ / σx̅ (Z- statistic)

In our example where the population distribution is normal with known

population mean and SD, we make our decision based on how large the
z-statistic we got. That is, the z-statistic is used to test the hypotheses.
Therefore, it is a test-statistic. The statistical test based on z-statistic is
called the Z test.

x̅ - μ is the observed difference between sample mean and the

hypothesized population mean; σx̅ quantifies sampling error, the
average distance between sample means and the population mean. It
measures the expected difference due to chance.

z-statistic is a test statistic used to determine whether the observed

difference is more than would be expected by chance alone.
 Hypothesis Testing
How large of z-statistic is large enough? Or how small the p-value is small
enough?
Decision rule 1:
The alpha level or the level of significance (α) : Alpha is a probability value
that is used to define the very unlikely sample outcomes if the null hypothesis
is true. It is determined before conducting the study.

Reject H0 if p≤α
Do not reject H0 if p>α

Commonly used alpha level: α =.05; α=.01; α=.001

For example, a decision rule with α=.05 defines an unlikely event as one that
occurs 5 times in 100 or less assuming H0 is true. This implies that we are
willing to make an error 5 times in 100 and reject the null hypothesis when it is
actually true.
 Hypothesis Testing
Decision rule 1:

In our example, suppose we set the alpha level to be .05.

Decision rule is:
Reject H0 if p≤.05

Do not reject H0 if p>.05

In this study, z=(29-26)/1=3, the p-value is prob(z≥3)=.0013,

which is less than .05, the alpha level. Thus, we reject the null
hypothesis.
 Hypothesis Testing
Decision Rule 2:
The critical region (reject region): The critical region consists of
outcomes that are very unlikely to be obtained if the null hypothesis is
true. The term very unlikely is defined by α.

The boundaries for the critical region are defined by α. In a normal

distribution, the boundary values are the z-values corresponding to
the α. And the z-values are called critical z-values.

Reject H0 if Z(observed) falls into the critical region

Do not Reject H0 if Z(observed) falls outside of the critical region
 Hypothesis Testing
Decision Rule 2:

In our example, the alpha level

is set to be .05. Identify the
critical z value beyond which
5% of the sample means fall.
From standard normal
distribution Table, Z(critical)
equals
1.65 for α=.05.

In this study, Z(observed)=

(29-26)/1 = 3
Z(observed)> Z(critical) i.e.,
Z(observed) falls in the critical
region. Thus reject the null
hypothesis.
 Hypothesis Testing
Example 2

The structure of a research study to determine whether prenatal alcohol

affects birth weight.
 Hypothesis Testing
H0: μ = 18

H1: μ ≠18
Non-directional test / two-tailed
test: no direction of the prediction
is specified in the statistical
hypotheses.

Our alpha level (α) is .05, for

non-directional test, the critical z-
values should be at both tails
corresponding to the p-value .025
in the normal unit table.
 Hypothesis Testing
Zx̅ = x̅ -μx̅ = -3
σx̅

Decision based on critical region: if we

set alpha level .05, the corresponding
critical z-values are ±1.96 (for a two-tailed
test). Because the z-statistic we
got here is -3, which is smaller than the
critical z = -1.96. In other words, it falls
in the critical region defined by alpha
α=.05, we reject the null hypothesis and
conclude that prenatal alcohol affects
birth weight.

Question: Test the hypothesis using Decision Rule 1 with p-value. P-value is
defined as the probability of getting something more extreme than your result
assuming H0 is true. What is p-value here?
 Hypothesis Testing

Steps in Hypothesis Testing of Mean Difference using test statistic:

Step 1: decide one-tailed or two-tailed test, set up hypotheses and the alpha
Level

Step 2: locate the critical region.

Step 3: Get the data and compute the test statistic.

Step 4: Make a decision.

 Hypothesis Testing

The locations of the critical region boundaries for three different levels of
significance: α = .05, α = .01, and α = .001 for non-directional test.
 Hypothesis Testing

Sample means that fall in the critical region (shaded areas) have a
probability less than alpha (p < α). H0 should be rejected. Sample means
that do not fall in the critical region have a probability greater than alpha (p
> α).
 Hypothesis Testing
Example 3: A researcher is using a sample of 16 laboratory rats to examine the
effect of a new diet drug. It is known that under regular circumstances these
rats eat an average of 10 grams of food each day. The distribution of food
consumption is normal with σ=4. The expected effect of the drug is to reduce
food consumption. The purpose of the experiment is to determine whether or
not the drug really works. The sample of 16 rats eat an average of 8 grams of
food each day.

Steps in Hypothesis Testing

Step 1: decide one-tailed or two-tailed test, set up hypotheses and the alpha
Level

Step 2: locate the critical region.

Step 3: Get the data and compute the test statistic.

Step 4: Make a decision.

 Hypothesis Testing

The distribution of sample means for n = 16 if H0 is true. The null hypothesis

states that the diet drug has no effect, so the population mean will be 10 or
larger.