You are on page 1of 7

What does Hypothesis Testing mean?

The immediate purpose is to determine the accuracy of hypothesis that the


investigator has collected a sample data, not a census.

How do we determine the accuracy of hypothesis?

1) one evaluate the accuracy of hypothesis by determining the statistical


likelihood or statistical methods that the data reveal “ True
Difference” not sampling error.
2) We evaluate the importance of a statistical significant difference by
measuring the significance of any change.

There are two approaches to Hypothesis Testing:


1) Sampling Theory Approach:
This approach is based on objective view of probability

The decision making is totally based on an analysis of


available sampling data

On the basis of this a hypothesis is established

Hypothesis accepted or rejected is based on the sample data collected

Bayesian Approach on Bayesian Statistics

It is centuries old and is referred to Bayes Theorem: Being used as an


alternative hypothesis testing procedures since the mid 1950

We may say that :


1) it is an extension of Classical Approach but goes beyond classical
approach, that it is also consider as the “Subjective Probability
Estimate”
2) the subjective estimate one based on the general experience rather
than on the data collected.
The Logic of Hypothesis Testing:

In classical tests of Significance two kinds of hypothesis are used:


1) the null hypothesis:
• It is a statement that no difference exists between the
parameters of the population, measurement of a sample and the
statistics ( is a measure from a recently drawn sample of the
population ) being compared to it.
• The investigator usually tests, to determine whether there has
been no change in the population or whether a real difference
exists.

1. Why not state the Hypothesis in Positive form?


2. Why not state that any difference between the sample
statistics and The population parameter; is due to same
reasons.
3. The positive hypothesis can never take valid grounds
for accepting hypothesis.

Logic behind the Null Hypothesis:

1. No to reject rather than accept the null hypothesis


2. It is argued that Null hypothesis is never be proved and therefore can
not be accepted.
3. statistical Testing only give a chance to
• Reject
• Fail to reject the hypothesis
• Accept means fail to reject

Statistical significance:

Following the “Sampling Theory Approach” we accept or reject a


hypothesis on the basis of sampling information. Since any sample will
almost vary, somewhat from its population. We must measure whether these
differences are statistically significant or insignificant.

A difference has “Statistical significance”. If there is good reason to believe


“ the difference does not represent random sampling fluctuations.”
CASE STUDY
Problem Statement:

A Company noticed a possible slow down in payments by the company’s


customers.

1. Generally the company has maintained an average age of 50 days ,


with a “standard deviation” of 10 days.
2. The company measures the “rate of payment” in terms of the “average
age of receivables outstanding”.
3. Suppose the controller has all of the customer’s account analyzed and
finds the average is now 51 days.

Is the difference statistically significant or insignificant?

It is a fact that the population average has moved from 50 to 51. while it is
of statistical significance. Whether it is of practical significance is another
question.
“If the controller judges, that this variation has no real importance, then it is
of little practical significance.

This example can be further explained to show how these concepts are used
to test statistical significance.

1. the null hypothesis states that the population parameter of 50 days has
not changed.
2. a second alternative hypothesis holds, that there has been a change in
average days outstanding: the sample statistics of 54 indicates the
population value probability is no longer 50.
3. the alternative hypothesis is the logical opposite of the null
hypothesis.
4. the null hypothesis H0 is : there has been no change from the 50 days
average age of account outstanding.
5. the alternative hypothesis HA may take several forms. The alternate
may take the form of “not the same” or “ greater than” or “less than”
form.
6. the average age of accounts has changed from 50 days.
7. the average age of receiving outstanding accounts has increased
( decreased from 50 days.

Methodology to test these kinds of tests:


Two Tail Test :
• It is non directional test: tests two possibilities.
• One possibility is “the average could be more than 50 days or it could
be less than 50 days.
• To test this hypothesis “the regions of rejection are divided in to two
tails of the distribution”.
One Tail Test:
• It is one directional Test:
• It places the entire responsibility of an unlikely outcome into the tail
specified by the alternative hypothesis.
• Hypothesis for example may be expressed as
Null Ho : µ = 50 days
Alternative HA : µ ≠ 50 days ( not the same case)
Or
HA : µ > 50 days ( greater than case)
Or HA : µ < 50 days ( less than case)

How to take decision:

• Take no corrective actions if the analysis shows that one can not reject
the null hypothesis.
• Not to reject rather than accept the null hypothesis
• It is also argued that null hypothesis can never be proved and
therefore can not be accepted.
• Statistical testing only gives a chance to
a) reject,
b) fail to reject the hypothesis
• it is also common in research circles that Accept the null hypothesis
rather than fail to reject.

Type I Error
( α ) is committed when a null hypothesis is rejected. The α value is called
the level of significance and is the probability of rejecting the null
hypothesis.
Type II Error
Β is when one fails to reject the false null hypothesis.

• The hypothesis testing places a greater emphasis on Type I error.


• If we reject a null hypothesis ( finding a significant statistical
differences) then we are accepting alternative hypothesis.
• In either case accepting or rejecting a null hypothesis we can make in
corrective decisions.
• A null hypothesis can be accepted when it should have been rejected,
when it should have been accepted.
• Two decisions can be made in this regard.
“One may accept or reject the null
hypothesis, thereby accepting the
alternative”
Two of these situations result correct decision, the other two
lead to decision error.
Statistical Significance
Our difficulty with working on data generated from a sample is that
“whether our finding will generalize the population from which the data is
collected.”
Another fact is
“That a difference between the population and
the sample has occurred even the probability
sampling procedure have been followed.”
Or
“There is always the possibility that
sampling error has occurred”.

If this happens the sample will not generalize the population and finding will
be invalid, then what to do to prove
“the confidence level that we provide an indication that our findings can be
generalized to the population. If we are not certain about that , we need
statistical techniques to establish that findings exists in the population and
how much risk we can take to infer that our findings based on the sample
will be found in the population .

These two elements that


1. how confident we are, that our findings can be generalized?
2. how much risk can we take to infer this.
The more important is that statistical significance can be employed only in
relation to samples that have been drawn using probability sample.

All the statistical Tests have a common structure

1. Set up a null hypothesis.


This refers that two variables are not related in the population. For
example in the case study or gym visitors, there is no relationship
between genders and visiting the gym in the population from which
the sample was drawn.
2. establish the level of statistical significance that you find better or
acceptable.
a) A measure of degree of risk:
That you might reject the null hypothesis explaining that there
is a relationship in the population, when you should have
supported it that there is no relationship in the population.

b) Level of statistical significance:


Are expressed as probability levels that is the probability of
rejecting the null hypothesis when you should be confirm right.
c) The convention among social researchers:
The maximum level of significance that is accepted is P < 0.05
which explains that there are five chances in hundred that you
could have a sample that shows a relationship existence.( you
would reject the null hypothesis.

3. determine the significance of your findings ( use a statistical test like


Chi- Square)
4. there are two types of error that can be made when we use statistical
significant techniques.