Confidence intervals

Question 1

Generalisation

Answer 1

-we can generalise the results of our study to the population only if the sample is reasonably representative of the population

Question 2

Confidence intervals

Answer 2

-help see how close the approximation of a measure in a sample is to the population

Question 3

95% confidence interval

Answer 3

-we are 95% certain that the true population mean lies between these two numbers

Question 4

Degree of confidence

Answer 4

• usually 95% confidence degree is reported
• this is derived from the idea that the probability of type 1 error (alpha) is 5%
• a higher degree of confidence can be set to 99% but a wider interval will be seen

Question 5

The width of the interval

Answer 5

• a wide interval at a fixed degree of confidence indicates that the estimate is not precise
• narrow interval is a precise estimate
• width of interval depends on the size of the standard error (variability) which depends on sample size

Question 6

Capturing the value of no difference

Answer 6

• if the 95% CI crosses 0 point for the difference between means then the results are statistically not significant
• if it crosses value 1 for ratio measures such as OR then it is not significant
• if it crosses infinity for inverse ratios such as NNT then it is not significant
• this is value of no difference

Question 7

Value of no difference

Answer 7

• 0 for means
• 1 for ratios
• infinity for NNTs

Question 8

How can you reduce the width of the confidence interval?

Answer 8

• choose a smaller degree of confidence level e.g 90% rather than 95%
• reduce standard deviation by using stratification approach
• take larger samples to increase N

Question 9

What do confidence intervals inform us about?

Answer 9

1. degree of confidence in our sample
2. precision of a result
3. clinical significance (magnitude of an effect)
4. statistical significance