Hypothesis Testing

Question 0

Procedure for hypothesis testing

Answer 0

Ask question about population parameter
Translate into null hypothesis 
Pick significant level α 
Collect data
Calculate test statistic 
Convert to p value
Make decision
Conclusion

Question 1

Hypothesis testing

Answer 1

An objective framework for making scientific conclusions based on a sample of data; uses proofs by contradiction (innocent until proven guilty)

Question 2

Null hypothesis (Ho)

Answer 2

The hypothesis to be tested (usually of no difference, no effect, or no association between a risk factor and disease)

Question 3

Alternative hypothesis (H1)

Answer 3

The hypothesis that contradicts the null hypothesis (usually the research hypothesis of interest)

Question 4

Significance level

Answer 4

Cut point used for choosing between H0 and H1
Represents the probability for choosing H1 when H0 is really true (the probability of rejecting a true null hypothesis)
α= .05 implies that there is a 5% error, which means 5% of the time you conclude there is an association when there really isn’t one (false positive)

Question 5

P-value

Answer 5

Probability of obtaining the observed sample estimate (or a more extreme estimate) by chance alone if the null hypothesis is true. The p value is not the probability the H0 is true (a common error).
If p < α our data is unlikely if H0 is true. Therefore reject H0. Findings are statistically significant. If p > α our data is not unlikely if H0 is true. Therefore do not reject H0. The results are not statistically significant. Any observed differences are compatible with chance.

Question 6

Paired T test

Answer 6

T = d/SE

Paired samples- two groups of subjects individually matched

Question 7

2 sample independent t test

Answer 7

Done with sample of subjects is randomly assigned to receive either the drug or the placebo (only one treatment per subject)

Question 8

F-test

Answer 8

Extension of t test to comparison of more than two groups

Question 9

Statistical significance

Answer 9

Only the results (ex. Difference between treatments) were probably not due to chance. Statistically significance does not mean the results are clinically significant. Increasing sample size, decreases the SE in the t-test which increases t and decreases the p value. Lack of statistical significance means that there is no difference or the sample sizes were too small to be able to detect a difference