Scribd
Upload
9 views

Lecture7 Introduction to Hypothesis Testing

Uploaded by

Ricx Rosco
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
9 views

Lecture7 Introduction to Hypothesis Testing

Uploaded by

Ricx Rosco
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 19

Hypothesis Testing

•••••••••••••••••••••••••••••
•••
Introduction to Hypothesis
Testing
Research is designed to answer questions, and
statistics are the tools to help us reach answers and
make decisions. Most often, research questions are
addressed via hypothesis testing.
Hypothesis testing are procedures designed for
determining whether a hypothesis is supported by
data. To do this, we will use inferential statistics
that are used to draw conclusions about populations
based on data collected from samples.
Inferential statistics are built upon probabilities,
specifically determining the probability that the data
represents a particular population.
c
e
h
ta
e
• D
ts
e
Hypothesis Testing
te
is
h
Steps
s
ie
S
g
ti
1 n
a
n
p
t
tp
iy
2
h
rs
o
e
o
tu
3 p
ir
s
rc
tia
o
u
4 a
lw
d
tn
y
e
h
Imagine that a researcher wants to investigate
whether entrepreneurs who exercise daily have
higher sales revenue in their company. Average
sales revenue for a company is ₱100,000.
• State the
Step 1 hypothesis using
your own words
Hypotheses are reasoned predictions
about the outcomes of a study. It is important
that researchers not only consider what they
expect or hope to happen, but also all of the
possible outcomes of their researcher and data
analysis. Therefore, when stating research
hypotheses it is useful to state a set of
hypotheses that are both mutually exclusive
(complete independent, and only one can be
true) and exhaustive (no other possibilities
exist).
For our example, there are 3 possibilities:
• Those who exercise daily do not
1 have higher a difference in sales
revenue in their company.
• Those who exercise daily have
2 lower sales revenue in their
company.

• Those who exercise daily have


3 higher sales revenue in their
company.
Step 2 • Design the Study

Once all of the possible


hypotheses/outcomes have been formalized, a
proper research study can be designed to
effectively “test” the hypotheses. The topic of
research design is not the focus in this course,
but I will briefly emphasize the importance of
the quality of the statistical analysis, which is
greatly dependent upon the quality of the
research study from which the data
originated. In other words, statistics cannot
make up for poorly conducted research.
Regarding our example, let’s imagine
that a researcher designs to obtain a
random sample of entrepreneurs who
exercise daily to determine if such
individuals have higher than average
sales revenue. Because it is now known
that the average sales revenue for a
company is ₱100,000, evidence
supporting the researchers prediction
would be if the entrepreneurs who
exercised average in revenue greater
than ₱100,000.
• Create Statistical
Step 3 Hypotheses

Statistical hypotheses are based


upon the written hypotheses
previously stated in step 1, but
describe the population parameters of
interest using statistical symbols.
More specifically, there are two types
of statistical hypotheses: the null and
alternative.
Null Hypothesis
• symbolized as H0
• the hypothesis that no difference or no
difference of interest and thus always
includes an equal sign
Alternative Hypothesis
• symbolized as HA
• the hypothesis that corresponds to the
researchers prediction (In this sense,
the alternative hypothesis is often called
the research hypothesis.)
Alternative hypotheses can be either one-tailed or
two-tailed.
One-tailed alternative
hypotheses
• are set forth when the research
predictions the direction of the
expected difference (up or down)
Two-tailed alternative
hypotheses
• the researcher does not predict the
direction of the difference but simply
predicts there is a difference
For our example, we indicated that the
researcher expects entrepreneurs who
exercise daily to perform BETTER in sales
revenue. Therefore, this situation calls for a
one-tail alternative hypothesis.
The null hypothesis would be that the
entrepreneurs who exercise daily do not
have higher sales revenue of ₱100,000 in
their company
The alternative hypothesis would be that
the entrepreneurs who exercise daily do
have higher sales revenue of ₱100,000 in
their company
• Select the
Step 4 appropriate
statistical test
Once the hypothesis are stated, the
study is designed, the statistical
hypotheses are stated, and the data are
collected it is time to analyze the data to
determine whether there is evidence for
the null or alternative hypothesis. Such
evidence is obtained through use of
inferential hypothesis tests. These
include z-tests, t-tests, Analysis of
Variance, and Regression.
The Logic of Hypothesis
Testing
It is important to understand that it is impossible to
statistically demonstrate or prove that something is
“true”, instead we use statistics to determine how
likely something is to be true. This is why probability
is important in statistics. In order to determine a
probability regarding how likely a hypothesis is to be
true, we must make an assumption. Specifically,
when conducting statistical hypothesis tests, we start
by assuming that the NULL is true, and then
determine how likely our data is to have occurred
given this assumption. Because we assume the null is
true, whenever we conduct a test we are testing the
null hypothesis.
The researcher then makes a
decision:
Reject the null: if the data are too
unlikely too occur when the null is
true
Fail to reject the null: if the data are
relatively likely to have occurred
when the null is true
Statistical Significance: p <
.05
You may have heard the phrase “p is less
than .05” in the context of statistical analysis.
P<.05 simply indicates that when the null
hypothesis is true, the probability of
obtaining the data are less than 5 out of
100. This criterion, p<.05 is the most
common one that researchers use to
determine when to reject the null
hypothesis.
Therefore, if P <.05, we conclude that it is
probably safe to reject the null hypothesis
and further conclude that there is a
statistically significant evidence in favor of
the alternative hypothesis.
P < .05 – reject the null

P > .05 – accept the null (or fail to


reject)
Errors and Outcomes

As stated, a researcher can conclude


only 1 of 2 things: either to reject the
null. However, when we consider the
accuracy of these decisions and the
possibility for error (making an
incorrect decision) there are two errors
that can occur.
Type I Errors Type II Errors
• occur when the • occur when the
researcher rejects researcher fails to
a true null reject a false null
hypothesis hypothesis
• Stated differently, • In other words,
this is when a the researcher
researcher should have
concludes that rejected the null
there is a hypothesis
statistically because it is not
significant result true, but failed to
but the conclusion do so.
was made in error
because it does not

You might also like