RAMESH DEBUR
Hypothesis testing
The Way Back
Formulate a research QuestionDevelop a research MethodologyCollected DataSorted Data
The Way Forward
IntroductionGENERATE A HYPOTHESIS
NULL HYPOTHESIS ALTERNATIVE HYPOTHESIS
Significance The p Value
Errors TYPE 1 TYPE 2
ACCEPT OR REJECT HYPOTHESISINFERENCE & PRESENTATION OF THE DATA
Hypothesis Testing
Scientific Hypothesis testing – A Deductive method of accepting or rejecting a hypothetical
statement
Definition
A hypothesis consists either of a suggested explanation for a phenomenonEg: Gastric Juices produces Hunger
or A reasoned proposal suggesting a possible correlation between multiple phenomena
Eg: People who smoke more cigarettes are at a higher risk of developing lung
cancer
Therefore…..
Hypothesis testing implies either accepting or rejecting a certain statement.
Generating a Hypothesis In Scientific Research the hypothesis is an offshoot of the research question
Eg: Do people who smoke more cigarettes increase their risk of developing Lung Cancer
The Logic of Hypothesis Testing
All hypothesis are false until proven true The farther away from falsification the truer is the hypothesis
A Hypothesis is NEVER a Fact. We accept a hypothesis as true until it is falsified
Eg: Columbus wants to discover a route to India Columbus Discovers America Every body he sees are called Indian – Hypothesis Later learns they are not Indian – Hypothesis falsified
The NULL Hypothesis
The exact opposite of the perceived effect or change
Eg: Gastric Juices DOES NOT cause Hunger Smoking DOES NOT Increase the risk of Developing Lung Cancer
The Alternate Hypothesis
The Exact opposite of the NULL Hypothesis
Called alternate because the falsification is the primary logic of hypothesis testing Gastric Juices DOES NOT cause Hunger Smoking DOES NOT Increase the risk of Developing Lung Cancer
Easier to Falsify things than prove facts?????
Disproving Null Vs Proving the Alternate
Null Alternate
THE NEXT QUESTION
WHEN DO WE REJECT THE NULL HYPOTHESIS
ANSWER : WHEN THE CHANCES OF IT BEING TRUE ARE VERY SLIM ( NON SIGNIFICANT)
PROBABILITY TESTING
Probability (prob·a·bil·i·ty)
Similar to Chance:Derived from the Noun Probable What is a probability : The chance of a event occurring at any given time
The likelihood of an event having a particular outcome
Eg: Flipping a coinAll probability is between 0 and 1
Flip a coin
Likelihood of getting only 1 head in 1 flip = 0.5 2 flips = 0.34 4flips = 0.24 10 = 0.01 100=0.00001
Likelihood of getting atleast 1 head in 1 flip = 0.5 2 flips = 0.66 4flips = 0.76 10 = 0.99 100=0.999999
Probability as related to hypothesis testing
The p value
The likelihood that the data collected is equal to or more extreme than the null hypothesis (logic: The Null hypothesis is the extreme value of an experiment)
Alternatively: The probability that the expected outcome occurred purely by chance
Significance - Definition
The significance level of a test is the probability that the test statistic will reject the null hypothesis when the [hypothesis] is true.
IN REAL TERMS
Get Test statistic (outcome of research )
Null hypothesis – test statistic if > chance : accept test and reject null
If ≤ chance : reject testGenerally the significance is kept at 0.05 or 1 chance in 100 or 0.001( 1 in 1000)
Errors of Hypothesis testing
Stastical Decision
True state of the Null Hypothesis
True HO False HO
Reject HO Type 1 ( α) Correct
Do Not Reject HO Correct Type II (β)
α is also called as the significance value and is determined by the investigator
A type 1 error is considered much more serious than a type 2 error
Why? EG. A new drug is introduced into the market which can potentially cure hypertension but can also cause sudden death. Evaluate the chances of sudden death by the drug
Statistical Decision
True state of the Null Hypothesis
True HO NO Death False HO Death
Reject HO Type 1 ( α) (Accept drug)
Correct (Reject drug)
Do Not Reject HO Correct (Accept Drug)
Type II (β) (Reject Drug)
Top Related