Hypothesis Testing and P Value
BY DR ZAHID KHAN
SENIOR LECTURER KING FAISAL UNIVERSITY, KSA
Two ways to learn
about a population


Confidence intervals



Hypothesis testing
HYPOTHESIS
What do you mean by a Hypothesis?
A hypothesis is a proposition that is –


assumed as a premise in an argument / claim



set forth as an explanation for the occurrence of
some specified group of phenomena



A hypothesis is a prediction about the outcome of an
experiment. In market research this could be the
result of the out come of a focus or field study
Why do we make hypotheses?


The practice of science traditionally involves
formulating and testing hypotheses



Hypotheses are assertions that are capable of being
proven false using a test of observed data



Hypothesis testing is a procedure through which
sample data is used to evaluate the credibility of a
hypothesis
TYPES OF HYPOTHESIS
Null Hypothesis



The null hypothesis typically corresponds to a
general or default position
Making this assertion will make no difference and
hence cannot be proven positively

Alternate Hypothesis
 An alternate hypothesis asserts a rival relationship
between the phenomena measured by the null
hypothesis
 It need not be a logical negation of the null
hypothesis as it only helps in rejecting or not
rejecting the null hypothesis
Dependant and independent
variables


Shoppers in a store playing music shop spend
more.



Independent Variable:
 Music



in the store

Dependent Variable:
 Amount

spent in store
Example -- Continued
1.

Obtain a random sample of shoppers who go
to stores with music

2.

Check shop spending

3.

Compare sample data to hypothesis

4.

Make decision:
1.

Reject the hypothesis

2.

Fail to reject the hypothesis
TYPES OF ERRORS
What are errors in Hypothesis
Testing?
The purpose of Hypothesis Testing is to reject or not
reject the Null Hypothesis based on statistical
evidence
Hypothesis Testing is said to have resulted in an error
when the decision regarding treatment of the Null
Hypothesis is wrong
TYPES OF ERRORS
Actual State of Affairs
Belief

Decision

H0 is True

H0 is False

H0 is False

Reject H0

Type I Error
False Positive

Correct Rejection
1Power

H0 is True

Fail to Reject H0

Correct Failure to
Reject
1-

Type II Error
False Negative
Statistical Power
1.

Probability that the test will correctly reject a
false null hypothesis.

2.

When a treatment effect exists
1.

A study may fail to discover it (Type II error, fail to
reject a false null hypothesis)

2.

A study may discover it (reject a false null
hypothesis)
α, β AND THE INTER-RELATIONSHIP
During the Hypothesis Testing,
α – is the probability of occurrence of a Type-I Error
β – is the probability of occurrence of a Type-II Error

Relationship between α and β




For a fixed sample size, the lower we set value of
α, the higher is the value of β and vice-versa
In many cases, it is difficult or almost impossible to
calculate the value of β and hence we usually
set only α
INTERPRETING RESULTS
Interpreting the weight of evidence against the Null
Hypothesis for rejecting / not rejecting Ho
If the p-value for testing Ho is less than –


< 0.05, we have strong evidence that Ho is false



< 0.01, we have very strong evidence that Ho is false



< 0.001, we have extremely strong evidence that Ho is false



P value is taken as 0.05 or 5% because it is a standard icon & it
nearly corresponds to the difference of two standard errors.
Jury’s Decision
Did Not Commit Crime

Committed Crime

Guilty

Type I Error
Convict Innocent
Person

Correct Verdict
Convict Guilty
Person

Not Guilty

Correct Acquittal
Type II Error
Fail to Convict Innocent Fail to Convict
Person
Guilty Person
Level of Significance
1. Alpha: probability of committing a Type I
error
1.

Reject H0 although it is true

2.

Symbolized by

2. Obtained result attributed to:
1.
2.

Real effect (reject H0)
Chance
One Sided & Two Sided Tests


Consider two means A & B.



One sided test only tells you that A > B.



Two sided tests tells you that either A>B or A <B so leaving you with
two options.



Mostly Two sided tests are used except in cases of equivalence tests
like Lumpectomy done for Breast surgery as well as radical
Mastectomy.



One sided test would be whether Lumpectomy is worst for survival
than Radical Mastectomy and we don't bother about better survival
results.
Any Questions !!!!

Thank

You.

Hypothesis testing and p values 06

  • 1.
    Hypothesis Testing andP Value BY DR ZAHID KHAN SENIOR LECTURER KING FAISAL UNIVERSITY, KSA
  • 2.
    Two ways tolearn about a population  Confidence intervals  Hypothesis testing
  • 3.
    HYPOTHESIS What do youmean by a Hypothesis? A hypothesis is a proposition that is –  assumed as a premise in an argument / claim  set forth as an explanation for the occurrence of some specified group of phenomena  A hypothesis is a prediction about the outcome of an experiment. In market research this could be the result of the out come of a focus or field study
  • 4.
    Why do wemake hypotheses?  The practice of science traditionally involves formulating and testing hypotheses  Hypotheses are assertions that are capable of being proven false using a test of observed data  Hypothesis testing is a procedure through which sample data is used to evaluate the credibility of a hypothesis
  • 5.
    TYPES OF HYPOTHESIS NullHypothesis   The null hypothesis typically corresponds to a general or default position Making this assertion will make no difference and hence cannot be proven positively Alternate Hypothesis  An alternate hypothesis asserts a rival relationship between the phenomena measured by the null hypothesis  It need not be a logical negation of the null hypothesis as it only helps in rejecting or not rejecting the null hypothesis
  • 6.
    Dependant and independent variables  Shoppersin a store playing music shop spend more.  Independent Variable:  Music  in the store Dependent Variable:  Amount spent in store
  • 7.
    Example -- Continued 1. Obtaina random sample of shoppers who go to stores with music 2. Check shop spending 3. Compare sample data to hypothesis 4. Make decision: 1. Reject the hypothesis 2. Fail to reject the hypothesis
  • 8.
    TYPES OF ERRORS Whatare errors in Hypothesis Testing? The purpose of Hypothesis Testing is to reject or not reject the Null Hypothesis based on statistical evidence Hypothesis Testing is said to have resulted in an error when the decision regarding treatment of the Null Hypothesis is wrong
  • 9.
    TYPES OF ERRORS ActualState of Affairs Belief Decision H0 is True H0 is False H0 is False Reject H0 Type I Error False Positive Correct Rejection 1Power H0 is True Fail to Reject H0 Correct Failure to Reject 1- Type II Error False Negative
  • 10.
    Statistical Power 1. Probability thatthe test will correctly reject a false null hypothesis. 2. When a treatment effect exists 1. A study may fail to discover it (Type II error, fail to reject a false null hypothesis) 2. A study may discover it (reject a false null hypothesis)
  • 11.
    α, β ANDTHE INTER-RELATIONSHIP During the Hypothesis Testing, α – is the probability of occurrence of a Type-I Error β – is the probability of occurrence of a Type-II Error Relationship between α and β   For a fixed sample size, the lower we set value of α, the higher is the value of β and vice-versa In many cases, it is difficult or almost impossible to calculate the value of β and hence we usually set only α
  • 12.
    INTERPRETING RESULTS Interpreting theweight of evidence against the Null Hypothesis for rejecting / not rejecting Ho If the p-value for testing Ho is less than –  < 0.05, we have strong evidence that Ho is false  < 0.01, we have very strong evidence that Ho is false  < 0.001, we have extremely strong evidence that Ho is false  P value is taken as 0.05 or 5% because it is a standard icon & it nearly corresponds to the difference of two standard errors.
  • 13.
    Jury’s Decision Did NotCommit Crime Committed Crime Guilty Type I Error Convict Innocent Person Correct Verdict Convict Guilty Person Not Guilty Correct Acquittal Type II Error Fail to Convict Innocent Fail to Convict Person Guilty Person
  • 14.
    Level of Significance 1.Alpha: probability of committing a Type I error 1. Reject H0 although it is true 2. Symbolized by 2. Obtained result attributed to: 1. 2. Real effect (reject H0) Chance
  • 15.
    One Sided &Two Sided Tests  Consider two means A & B.  One sided test only tells you that A > B.  Two sided tests tells you that either A>B or A <B so leaving you with two options.  Mostly Two sided tests are used except in cases of equivalence tests like Lumpectomy done for Breast surgery as well as radical Mastectomy.  One sided test would be whether Lumpectomy is worst for survival than Radical Mastectomy and we don't bother about better survival results.
  • 16.