Stats 2 Flashcards
1) What does bivariate analysis mean?
1) looking at relationships between TWO variables
2) What is the regression coefficient
2) y=a+bx. b is regression coefficient, measure of the gradient
3) A regression coefficient of 0 means?
accept null hypothesis
4) What is correlation ?
4) measures the degree of linear association, between -1 and +1
5) What does a correlation coefficient of -0.8 mean
5) strong negative correlation
6) One positive of regression coefficient over correlation coefficient
6) can directly predict the outcome given the exposure
When would you use a correlation coefficient
7) when outcome and exposure cant be distinguished – one doesn’t cause other one e.g. weight and height
8) Give an example of descriptive analysis
8) mean, median, standard deviation
9) When do you need more complicated analysis
9) when you’re comparing more than two groups
10) What is ANOVA
10) analysis of the variance – used when theres more than two groups (extension of t-test)
11) When do you use multiple logistic regression and multiple linear regression
11) when theres more than two variables. Logistic when the outcomes are binary and linear when the outcomes are continuous.
12) Describe adjusting for cofounders
12) Result of study is effected by result of relating factor. e.g. testing for coffee and lung cancer (lots of coffee drinkers smoke so this could effect results)
13) What form does logistic regression give its output in and why ?
odds ratios as is categorical
14) Describe type one error
14) false rejection of true null hypothesis (false positive)
15) Describe type two error
15) wrong acceptance of false null hypothesis
16) Why does type 2 error often occur
16) inadequate sample size . Not enough power for stat sig
17) What is r(squared)
17) percentage of variation in outcome which can be accounted for by exposure variable
18) How do you work out sample size and power
18) need to know variance and SD and what clinically significant effect size you’re looking for. Put into the formula and tells you power
19) When would you lower the p value and why
19) if you do loads of stats test by chance 5/100 of them would be statistical significant so need to lower p incase occurs by chance – multiple outcome measures
20) Why is multiple subgroup analysis bad
20) if you have lots of subgroups then more likely to get a statistically significant result as doing more separate tests
21) what is the secondary hypothesis often related to
21) comparing subgroups
22) what is measure of effect size also known as
odds ratio