statistics 2

Question 1

what is observational studies?

Answer 1

record outcomes as they have
occurred, and there is no attempt to modify the frequency
of their occurrence. Examples include cohort, casecontrol, cross-sectional and ecological studies

Question 2

what is experimental studies?

Answer 2

Experimental studies involve investigators introducing or
removing an exposure (e.g. treatment or intervention) to
test relationships between exposures and outcomes.
Investigators have direct control over the study
conditions and exposure status. Examples include clinical
trials, many pre-clinical experiments.

Question 3

why is sample size important?

Answer 3

The statistical design of a study should be matched to the
objectives of:
Optimising resources (which could be scarce and/or
expensive)
Getting accurate and precise results
• Too small a study:
We may miss worthwhile effects
• Too large a study:
Wastes resources

Question 4

what is replication

Answer 4

Replication defines how many units receive the same exposure level

Question 5

what does improving the replication cause?

Answer 5

Increasing the replication improves the accuracy of
results and is essential to provide estimates of precision.
It generally increases the power of the experiment to
detect important differences between exposure levels

Question 6

what is blocking

Answer 6

Blocking is used to take account of differences (or
heterogeneities) in the units (e.g. male/female, old/young
patients)

Use of blocking can reduce variability which improves
precision

Question 7

what is randomisation and what is its advantages?

Answer 7

Some degree of randomisation is vital for the results of
an experiment to be meaningful.

For example, in a medical experiment where a drug is
being compared with a placebo, if randomisation is
not used, doctors may be tempted to give the drug to
the most seriously ill patients to try and help them
• This will bias the results
• Randomisation allows a fair comparison to be made
between exposures

Question 8

what is null hypothesis?

Answer 8

Null hypothesis is denoted H0
• Usually a precise statement about the population
parameters
• In the example, H0
: m = m0 = 76
• We will examine the data to see if there is evidence to
reject the null hypothesis.

Question 9

what is h1 hypothesis

Answer 9

Alternative hypothesis is denoted H1
• A (usually vague) statement of what will be accepted if the
data do not support H0
• E.g. “There is a difference between the population mean
and the hypothetical value”.

Question 10

what is type I error

Answer 10

A Type I error is the error we make if we reject the null

hypothesis when it is true

Question 11

what is type II error

Answer 11

A Type II error is an error we make, if we do not reject

the null hypothesis when, in fact, it is false

Question 12

if p value is smaller what does it mean about the null hypothesis?

Answer 12

• The smaller the p-value, the stronger the evidence
against the null hypothesis, i.e. there is a very small
chance that we would obtain a test statistic as or a more
extreme than ours if the null hypothesis was true.

Question 13

what should be dine to the null hypothesis if the significance level is 5%, and p < 0.05?

Answer 13

If the significance level is 5%, and p < 0.05, then the
result is declared significant and we reject the null
hypothesis, in favour of the alternative being true.

Question 14

what should be done to the null hypothesis when the p-value is not small

Answer 14

If the p-value is not small then we have insufficient
evidence to suggest that the null hypothesis is not true,
we can not reject the null hypothesis.
• Important: a non-significant result does NOT mean
that we can accept the null hypothesis.