Research Studies - Analysis of Data

Question 1

Confidence intervals
-what is this
-how is it calculated

Answer 1

A range of values within which the true effect of intervention is likely to lie
If at 95%, the true effect of the intervention should lie within this range 95% of the time

Standard error of the mean = SD/square root of n
-as n increases, SEM decreases

Lower limit = mean - (1.96 x SEM)
Upper limit = mean + (1.96 x SEM)

Question 2

Correlation
-what is this
-interpretation of r

Answer 2

How closely the points lie to a line drawn through the plotted data

r = 1 strong positive correlation
r = 0 no correlation
r = -r strong negative correlation

No info on cause and effect
No info on how much y will change if x is changed

Question 3

Linear regression
-what is this and how does it differ from correlation

Answer 3

Can predict how y changes if x changes

Question 4

Forest plots
-what is this used for

Answer 4

Graphical display of multiple results from different studies in a meta analysis

Shows
-CI
-mean
-line of no effect - studies that cross this line are insignificant
-diamond - result of the meta analysis

Question 5

Funnel plot
-what is this used for

Answer 5

Demonstrate the existence of publication bias in meta analyses

Question 6

Box and whisker plot
-what is this used for

Answer 6

Graphical representation of the sample minimum, lower quartile, median, upper quartile and sample maximum

Question 7

Histogram
-what is this used for

Answer 7

Graphical display of continuous data where values have been categorised into multiple categories

Question 8

Scatter plot
-what is this used for

Answer 8

Graphical representation using Cartesian coordinates to display values for 2 variables for a set of data

Question 9

Kaplan-Meier survival plot

Answer 9

Plot estimating the survival function showing decreasing survival with time

Question 10

Intention to treat analysis
-what is this

Answer 10

All patients randomly assigned to 1 treatment are analysed together, regardless of whether they completed the treatment

Avoid effects of crossover and drop out which may affect the randomisation to the treatment groups

Question 11

Distributions
-what is normal distribution and the percentages

Answer 11

Symmetrical - mean = mode = median

1SD - 68.3% of values

1.96SD - 95% of values

2SD - 95.4% of values
3SD - 99.7% of values

SD = square root of variance
Variance = measure of spread of scores away from the mean

Question 12

Positive skew and negative skew
-distribution of mean, median, mode

Answer 12

Positive skew
mean > median > mode

Negative skew
mean < median < mode

Question 13

Significance test types
-definition of parametric vs non-parametric

Answer 13

Parametric - can be measured
Non-parametric - cannot be measured

Parametric
-Student’s t test - paired, unpaired
-Pearsons - correlation of linear data

Non-parametric
-Mann-Whitney U - compares ordinal, interval or ratio scales of unpaired data
-Wilcoxon signed-rank - 2 observations on 1 sample after an intervention
-Chi-squared - compare proportions, percentages after an intervention
-Spearman - correlation of non linear data

Question 14

Significance tests
-H0, H1
-p value
-type 1, 2 errors
-power

Answer 14

H0 - 2 treatments are equally effective
H1 - there is a difference between the 2 treatments
p - probability of obtaining getting this result by chance if H0 is true

T1 error - H0 rejected when true (a)
T2 error - H0 accepted when false (b)

Power - probability of correctly rejecting H0
1-b
Can be increased by increasing sample size

Question 15

Number needed to treat
-what is this
-how to calculate it

Absolute risk reduction
-what is it
-how to calculate it

Answer 15

How many patients would require intervention to reduce the expected number of outcomes by 1

1/absolute risk reduction

Absolute risk reduction - absolute difference between the control event rate and experimental event rate

Question 16

Probability vs odds ratio

Answer 16

Probability - the fraction of times you’d expect to get a specific event after multiple trials

Odds - number of positive outcomes / number of negative outcomes

Odds ratio - odds of getting a specific event / odds of not getting the specific event

Question 17

Relative risk
-what is this
-how to interpret this ratio
-how to calculate the relative risk reduction

Answer 17

Ratio of risk in experimental group / risk in control group

1+ - rate of event is increased compared to controls
-1 - rate of event is decreased compared to controls

Relative risk change - absolute risk / control event rate