Power calculation and effect size

Question 1

SIGNIFICANCE LEVEL AND POWER

Answer 1

 Significance level (α) (0.05). The chance of saying there is an relationship/effect when in fact there is not.
 type 1 error (α) (e.g. 0.05): (5% risk of concluding that
a difference exists when there is no actual difference)

 Power (1‐β) e.g. 0.80. Statistical power is the likelihood that a study will detect an effect when there is an effect there to be detected.
 type 2 error (e.g. 0.20): (i.e. missing it 20% of the time)

 Type 1 error: False positive
The study showed an effect which in reality does not exist.

 Type 2 error (β): False negative
An effect was there but the study missed it.

Question 2

STATISTICAL ERRORS

Answer 2

Type 1: False positive
The study showed an effect which in reality does not exist.
Type 2: False negative
An effect was there but the study missed it.

Question 3

SAMPLE SIZE:

FACTOR 1: DEGREE OF CONFIDENCE

Answer 3

 Suppose that in a random sample of 100 people, the
mean number of times people reported feeling unwell
per year =40 (S.D.=10).
 For a given sample size, if you want to know the true
population mean, make the confidence interval wider.
 For the population, the true mean is:
95% C.I. 36.8‐43.2
99% C.I. 32.9‐47.1

Question 4

SAMPLE SIZE:

FACTOR 2 - NUMBER OF DATA POINTS

Answer 4

With larger samples, confidence intervals are narrower.
 For 100 people, mean = 40 (SD=10), 95% C.I.=38‐42
 For 1000 people, mean = 40 (SD=10), 95% C.I.=39‐41

Question 5

SAMPLE SIZE: FACTOR 3 -

SCATTER OF DATA

Answer 5

If the data are more scattered the confidence interval is wider. E.g. for 100 people:
 Mean = 40, SD=10, 95% C.I.=38‐42
 Mean = 40, SD=40, 95% C.I.=32‐47

Question 6

SAMPLE SIZE IN QUALITATIVE STUDIES

Answer 6

 No definite rules to be followed.
 It will depend on what you want to know, the purpose of the inquiry, what is at stake, what will be useful, what will have credibility and what can be done with available time and resources.
 With fixed resources which is always the case, you
can choose to study one specific phenomenon in depth with a smaller sample size or a bigger sample size when seeking breadth.

Question 7

NON SAMPLING ERROR

Answer 7

 Errors may result from:
 The manner in which the response is elicited
 The social desirability of the response

Question 8

SOME SOURCES OF BIAS in

epidemiological studies

Answer 8

• Measurement error: exposure, outcome, confounders
• Loss to follow ‐up in longitudinal studies and selection bias in case control studies
 Selection bias occurs when there is selective (differential) referral or enrollment of cases or controls, and the selection factor is related to the probability of being exposed.