Biostatistics
Biostatistics
org
Julia Saperia
European Medicines Agency
percentage
spread
mean
measurement
information
graphs
confidence
variability
data
average
evidence
probability median
2
An example
3
An example
4
A sillier example
• So the means are the quite similar…but the two groups look
different
5
A sillier example
6
The histogram
7
Picturing “normal” data
~99%,
~95,5%
68,3%
8
Biostatistics
9
(a population is a set of all
Population conceivable observations of a
certain phenomenon)
Inferential statistics
11
Descriptive statistics (2)
12
Describing data
measure of precision of
estimate of population mean the sample mean
15
Confidence intervals
16
A real example
17
A real example
• Assume 5 degree lateral wedged insoles and flat control insoles reduce
pain to the same extent (this is our null hypothesis)
• Collect data in both groups of patients and calculate:
the mean reduction from baseline in the 5 deg lateral insole group (0.9)
the mean reduction from baseline in the flat insole group (1.2)
the difference between the two groups in mean change from baseline (0.9
– 1.2 = -0.3)
• We can see that there was some reduction in pain in both groups and
that the reduction was greater in the flat insole group
• The 95% confidence interval tells us how important that difference is
18
A real example
19
Statistical significance vs clinical significance
20
Quantifying statistical significance – the p-value
21
The p-value
22
Significance level
if p > 0.05 do not reject the null hypothesis conclude that there is not
sufficient information to reject the null hypothesis
(See Statistics notes: Absence of evidence is not evidence of absence:
BMJ 1995;311:485)
The authors of the insole paper did not cite the p-value for the result we looked
at but we can conclude that p>0.05 because the 95%CI crosses the null value
and so we cannot reject the null hypothesis (of no difference).
23
Statistical mistakes/errors
True state of nature
Hypothesis accepted H0 H1
Test decision
H0 Correct
1-
H1 False positive
Type I error:
Error of rejecting a null hypothesis when it is actually true
( error, error of the first kind)
represents a FALSE POSITIVE decision: A placebo is declared to be more
effective than another placebo!)
24
Statistical mistakes/errors
True state of nature
Hypothesis accepted H0 H1
Test decision
Type II error:
The alternative hypothesis is true, but the null hypothesis is erroneously
not rejected. (β error, error of the second kind)
FALSE NEGATIVE decision: an effective treatment could not be
significantly distinguished from placebo
25
Which sorts of error can occur when performing statistical tests?
Statistical mistakes/errors
True state of nature
Hypothesis accepted H0 H1
Test decision
POWER 1- : The power of a statistical test is the probability that the test will reject
a false null hypothesis.
26
Statistical mistakes/errors
True state of nature
Hypothesis accepted H0 H1
Test decision
Consumer‘s risk?
Producer‘s risk?
27
Compare a statistical test to a court of justice
Statistical tests vs court of law
True state of nature
H0 H1
Innocent Not Innocent
Decision of the court
Judged
False positive: A court finds a person guilty of a crime that they did not actually commit.
False negative: A court finds a person not guilty of a crime that they did commit.
28
Minimising statistical errors
Remember:
How do I gain adequate data?
Thorough planning of studies
• Defining acceptable levels of statistical error is key to the planning
of studies
• alpha (in clinical trials) is pre-defined by regulatory guidance
(usually)
• beta is not, but deciding on the power (1-beta) of the study is
crucial to enrolling sufficient patients
• the power of a study is usually chosen to be 80% or 90%
• conducting an “underpowered” study is not ethically acceptable
because you know in advance that your results will be inconclusive
29
Deciding how many patients to enrol
30
Conclusion
31
Further reading
• www.consort-statement.org
standards for reporting clinical trials in the
literature
• Statistical Principles for Clinical Trials ICH E9
useful glossary
• http://openwetware.org/wiki/BMJ_Statistics_Notes_series
coverage of a number of topics related to
statistics in clinical research, mostly by Douglas
Altman and Martin Bland
32