5] Hypothesis Testing 1 Flashcards
What is Occam’s Razor
It is when all else being equal is simpler, more economical explanations should be preferred over more complex ones
Fewer number of assumptions needed when attempting to explain some observation the better
E.g: 440k need to have kept the moon landing hoax a secret = highly improbable
The research process method
1] Review the literature
2] Define the research problem
3] Formulate hypothesis or objective
4] Design appropriate research method
5] Collect data
6] Analyse and interpret data
7] Report findings irrespective of results
What is hypothesis
A statement about a population parameter, subject to verification
Data is collected and tested by inferential statistics
What are the four general characteristics of a good hypothesis
1] Must be testable and falsifiable
2] Logical- informed by previous theories or logical reasoning
3] Positive- make a positive statement about the existence of a relationship or effect
4] Precise- bad hypothesis “males are bigger than females”
What is null hypothesis significance testing
It is the most common approach
We formulate two versions of our hypothesis a null and alternate
What is a null hypothesis
The state of affairs where no effect exists, there is no difference between groups and no relationship between variables
Why do we need a null hypothesis
We can never really prove a hypothesis but we can disprove it
We can either reject the null hypothesis or fail to reject it
What is the alternate hypothesis
It describes what you will conclude if you reject the null hypothesis.
It is accepted if the sample data provides us with enough evidence that the null is false
If there is a small effect between two groups how do we know when it is a genuine effect or due to chance
We start with assuming the null is accurate
Then test the probability that we would have found the effects we observed in our sample, if the null hypothesis was true for the population
If there was a high probability we would have found the effect in our sample due to chance we conclude the null is probably true. If its a low probability we conclude its probably false
How do we specify a level of probability that is low enough to conclude the null is false
By a significance level also called the alpha level
We typically use an alpha level of .05, which is less than 5% probability that the results (small effect) are due to chance
5 Steps of null hypothesis significance testing
1] State the null and alternate hypotheses
2] Specify the significance level (a)
3] Using your sample data, compute sample statistics
4] Compute the probability (p) that the observed value of the test statistic would occur if null was true
5] Compare p to a, accept or reject the null
What are the two major forms of measurement error
1] Systematic error
2] Random error
What is Systematic error
Errors are repeated the same way throughout an investigation
What is a Random error
Errors that are not systematic, usually due to sampling
They occur due to individual differences
How do we control measurement errors
1] Use large sample sizes for random errors
2] Use reliable and valid measurement tools for systematic errors