p.
42 = LENGTH-TIME bias = when screening test preferentially detect less agressive
form of disease and therefore increases survival time.
p59 = Clinical trials
Randomization minimizes the effect of confounding.
Single blinded study patients dont know if they take drug or placebo = minimize
placebo effect.
Blinding caregiver = less observer bias.
Intention-to-treat = principle usd in analysis. If a patient who is assigned to placebo
group begins to take drugs, or patient from study group stop taking their medicine
we still analyze them as they are still in their groups. (increase randomization, less
false positives but more false negatives)
as treated rule opposite, we analyze patients with that group they are similar to
(from placebo started to take drugs now analyze with treated population)
P.63 = Statistical distribution
Take value -> substract from it mean -> divide by SD = Z score ()indicate how many
SDs a given value is from the mean.
p.66 = Comparing Groups
Variables to be compared:
1. Means:
1.1. Independent samples = Two-sample T test (2 groups) or ANOVA (more
than 2 groups)
1.2. Same individuals followed over time = Paired T test.
2. Proportions:
1.1. Big sample size = Chi-square test
1.2. Small sample size = Fisher`s exact test!!! (if expected value in either of
the cell is less than 10)
Two-sample T test = Student test.
P.74 = Statistical power
Power of the study depends on:
-alpha level = lowering the alpha level (strenghtening the significance criterion, eg
not 0,05 but 0,01 we change criteria to more strict = if we have we surely know that
it works and not to chance alone) = but this decreases the power of the study (we
have less chances to see results, even though they are and increase risk of beta-
error (FN))
-the magnitude of difference in outcome between the study groups (subte
difference is more difficult to detect than a big difference)
-increasing the sample size increases the probability of detecting the difference in
outcome between the study groups