Stats 2

Question 1

What is the null and study hypothesis’s ?

Answer 1

Null= no dif between exposed and unexposed 
Study = there is a dif between exposed and unexposed

Question 2

What is a p-value?

Answer 2

The chance that what you saw happened by chance

P= 0.02 means probability it happened by chance is 2 in 100

Question 3

3 steps to hypothesis testing

Answer 3

1) specify the study hypothesis and the null hypothesis
2) assume the null hypothesis is true
3) calculate the p-value- if low (<0.05) reject null hypothesis

Question 4

What is a type 1 error?

Answer 4

False positive -reject the null hypothesis even though it is actually true

Question 5

What is a type 2 error?

Answer 5

False negative - fail to reject null hypothesis even though it is actually false.

Question 6

How do you calculate the T statistic?

Answer 6

Observed mean difference/

Standard error of difference between means

Question 7

What assumptions does the t-test make?

Answer 7

The outcome is continuous and normal distribution

And variance in the two groups is equal

Question 8

What does the levene’s test measure?

Answer 8

That variance between two groups is equal (one of the assumptions needed for the t-test)

Question 9

In spss what is the p-value called?

Answer 9

Sig

Question 10

Chi- squared tests are used to calculate p-values from what kind of data?

Answer 10

Categorical data

Question 11

T- tests are used to calculate the p-value from what kind of data?

Answer 11

Continuous data

Question 12

What is the student t-test?

Answer 12

Independent samples t-test

Used when u want to compare two groups of continuous, normally distributed variables

(Assumes that variance/scatter in two groups are similar)

Question 13

How does the t-test work?

Answer 13

Uses the observed difference in sample means and the standard error (sampling error) for the dif in means to calculate the p value

Question 14

What is this t-statistic?

Answer 14

Difference in sample means/ standard error of the dif in means

Question 15

What would result in a small t-statistic and what does it mean?

Answer 15

Either small difference in means or a large SE.

Means probability that the observed dif happened by chance.= large = large p-value

Question 16

What is a parametric test? Name one.

Answer 16

Only valid when the data is normally distributed and 2 populations have equal variance
Example= t-test

Question 17

What are the negatives and positives of NON- parametric tests?

Answer 17

Pos: makes no assumptions of underlying distribution of data.

Neg: less powerful than parametric test- difficult to get confidence intervals

Question 18

How to transform moderately positively skewed data?

Answer 18

Logarithm of each number

Question 19

How do you transform strongly positively skewed data?

Answer 19

Reciprocal (l/x)

Question 20

How do you transform weakly positively skewed data?

Answer 20

Square root

Question 21

How do you transform moderately negatively skewed data?

Answer 21

Square

Question 22

How do you transform strongly negatively skewed data?

Answer 22

Cube x3

Question 23

How do you transform data with unequal variation

Answer 23

Log, reciprocal or square root

Question 24

How to describe skewed variables if data not normally distributed?

Answer 24

1) present medians
2) present interquartile range
3) difference between two medians (CI not easy)

Question 25

Name two non-parametric equivalents of the t-test

Answer 25

Wilcoxon rank sum test | Mann-Whitney U test

Question 26

Describe wilcoxon rank sum test

Answer 26

Two independent groups (1 smaller) 1) rank all observations in asset order 2) sum ranks for smaller group (t-statistic) 3) look up t on wilcoxon rank sum table of critical values to get p-value

Question 27

What assumption does wilcoxon rank sum test and Mann-Whitney U test make?

Answer 27

That there are no tied ranks- if many identical results - complicated corrections are applied

Question 28

What assumptions does the Chi- squared test make?

Answer 28

Each subject contributes data to only one cell | The expected count in each cell should be at least 5

Question 29

What test can u use for p-value if u have a 2X2 table with numbers <5

Answer 29

Fishers exact test

Question 30

What is the chi-squared test for trends?

Answer 30

Used when exposure variable is ordered (eg age) and outcome is binary.

Question 31

What does the test statistic measure (chi-squared)

Answer 31

How close the observed values in a table are to expected values were there no true association

Question 32

How do you calculate the chi-squared statistic?

Answer 32

For each cell: subtract expected (E) from observed (O) then square and divide all by E Sum over all the cells to give overall chi squared statistic (x^2)

Question 33

A large chia squared statistic means the data is more or less consistent with the null hypothesis?

Answer 33

Less

Question 34

A large x^2 value means aargh or small p-value?

Answer 34

Small

Question 35

What is the difference between hypothesis testing and measures of effect?

Answer 35

Hypothesis testing = assesses the significance of an association Measure of effect (or/rr) assesses the magnitude of an association

Question 36

On a scatter plot comparing two continuous variables - what do you plot on x-axis?

Answer 36

Y axis is response or outcome and x axis is the exposure

Question 37

What does correlation measure?

Answer 37

The closeness or degree of association between two continuous variables

Question 38

What does Pearsons correlation coefficient (R) measure?

Answer 38

The degree of linear association

Question 39

Can pearsons coefficient be used in data that is unevenly distributed?

Answer 39

No. Only evenly distributed data

Question 40

Between what values will R always lie?

Answer 40

-1 and + 1

Question 41

What does an R of +1 or -1 look like?

Answer 41

R=1 is a positive straight line (going up) r=-1 is a straight line going down as x increases y decreases R= 0 is no correlation

Question 42

Can you calculate 95% CI for R?

Answer 42

Yes - tells us where we are 95% confident the true value lies

Question 43

What does r squared represent?

Answer 43

An estimate of the proportion of the total variation of our outcome variable that is explained by the other variable. R between blood pressure and stress =O.80, r2= 0.64. Therefore stress accounts for 64% of total variation in BP.

Question 44

Name two non-parametric correlation coefficient tests.

Answer 44

1)Spearman' s rank correlation coefficient 2) Kendall's coefficient Both rank the values of each of the two variables and examine how closely the ranks are correlated

Question 45

Can you use correlation when there are multiple measurements on each individual?

Answer 45

No

Question 46

How should you describe correlation?

Answer 46

Talk about the direction (pos or neg R), the strength (size or R) and the significance (p-value)

Question 47

In relations to calculating linear regression what does each symbol represent y= a + bx

Answer 47

``` Y= outcome variable X= exposure variable A= the point where line crosses the y axis (where x is 0) B= slope of the line A and b are the regression coefficients ```

Question 48

How do you determine he best fit line (linear regression)

Answer 48

Method of least squares ( find the regression coefficients a and b that minimise the vertical distances between the points and the line

Question 49

What is the most important value in linear regression (i.e. Y, x, a or b)

Answer 49

B is most important as is tells you how much the outcome y increases or decreases per unit increase in the exposure Can calculate 95% CI around b P- value to test null hypothesis (no association)

Question 50

What does linear regression tell you?

Answer 50

The nature of the relation between x and y by giving us the equation of the best fit line but does not tell us how well the line fits the data.