Statistics and methodology
Not your usual “Notes” - How to use this to ACTIVELY RECALL:
▪ Bullet-point Question/Answer format
▪ Based on Lectures from course AB_1201 provided by VU Amsterdam
▪ Colour the questions RED until you perfect them. Colour them GREEN once you’ve
mastered them. Then, ONLY practice the red question, until they’re all green.
▪ Follow the chapters from the book/lecture slides to view images and visuals
corresponding to the notes.
Recap notes RBMS
● RQ → hypotheses → study design & data collections → descriptive statistics → inferential
statistics → conclusion → RQ
● What is null hypothesis testing? We create a H0 where there is no effect/association. Then we
test P(data|H0) and if that is high, then we keep H0 (note that we don’t accept it as true. We
simply retain it). If low, we reject H0.
● What is an alternative hypothesis? There is an effect/an association
● How does a well formulated RQ look? Population, Intervention, Comparison, and Outcome.
● What is a P-value? A P-value is the probability of getting the observed data assuming the null
hypothesis is true. Low P-value (P<0.05)=
reject H0. High P-value (P=>0.05) = retain
H0.
● What is a two-tailed vs a one-tailed
hypothesis? Two tail hypothesis → about
an ‘effect’ or ‘difference’ caused by A on
B. Could be an increase or a decrease -
bidirectional. One tail hypothesis → more
specific of the direction. E.g. “smaller
than”, “greater than”, etc.
● What are the two kinds of biomedical
research designs? Observational and
experimental
● What are the kinds of observational
experimental designs?
Cross sectional = observational in the
moment study
Cohort/prospective study = follow into
the future
Case control = observational comparison case and control
● What are the kinds of experimental studies?
Randomised control trial = randomly assorted to 1 of the 2 conditions
Cross-over study = participants are in both conditions but with a washout period.
● What is the goal of descriptive statistics? To present, organise and summarise the data
observed in the sample.
, ● What are the descriptive statistics used? Frequency (or proportion as a %), mean, median,
mode, variance, range, IQR, STDEV + graphs and figures
● What are inferential statistics for? To draw conclusions about a population based on data
observed in a sample
● What is used in inferential statistics? Statistical tests that results in a p-value = probability of
the data given that the H0 is true.
Very unlikely (below 0.05) → reject H0. That means low p-value means there is
significance.
Hence, alpha = 0.05 (5%)
Statistical tests
● What are the 2 broad groups for
statistical tests? Test of association
(are two variables related) and test of
difference (Do two or more groups
differ → comparing proportions or
means)
● What are the tests of association?
○ Chi squared test for
independence → tells you if
there is an association
between 2 nominal variables
(e.g. smoking yes/no, lung
cancer yes/no)
○ Pearson’s R: measures
strength of linear relationship.
Assumes normal distribution.
○ Spearman’s Rho: used for
association when Data are not
normally distributed and
Relationship is monotonic but
not perfectly linear + data is
ordinal
“Does tumor stage (I–IV) correlate with survival time?”
○ Regression analysis: models how one variable predicts another. “How much does
cholesterol increase risk, controlling for age?”
● What are the tests of difference?
○ Binomial test/One sample Z test: Compare sample proportion to population
A vaccine trial expects 95% immunity, but your sample shows 90%. Is 90%
significantly different from 95%?
○ Chi squared goodness of fit test: Compare observed frequencies to expected
frequencies
○ Two-sample Z test: Comparing proportions between two independent groups.
Comparing infection rates: Hospital A: 8%, Hospital B: 5%
Is Hospital A significantly worse?
, ○ Chi squared test homogeneity: Comparing distribution of a categorical outcome
across groups
Do treatment groups differ in
side-effect type?
○ One sample T test -
comparing sample mean to
known reference
○ Independent samples T test -
comparing means of two
independent samples
○ Dependent samples T test -
comparing means in same
sample before and after
● What are the features of a normal distribution? Mean 0 and STDEV 1.
● What do Z scores help with? Basically a standardised version of an observation. Z scores
helps to compare two scores from two different distributions. Z score is basically the SD
away from the mean of each situation
● How does a t-distribution work? Basically you
have a STDEV of the sample, which you
assume is equal to the population’s STDEV.
You calculate a t value. There are 2 t-critical thresholds (t-critical-0.025 and t-critical-0.975)
on the normal curve that your calculated t-value has 95% chance of falling within. The
remaining 5% is for below 0.025 or above 0.975 which is very unlikely. So if your t-value
falls somewhere there, it means it’s very unlikely to get your results given H0 is true, and so
Not your usual “Notes” - How to use this to ACTIVELY RECALL:
▪ Bullet-point Question/Answer format
▪ Based on Lectures from course AB_1201 provided by VU Amsterdam
▪ Colour the questions RED until you perfect them. Colour them GREEN once you’ve
mastered them. Then, ONLY practice the red question, until they’re all green.
▪ Follow the chapters from the book/lecture slides to view images and visuals
corresponding to the notes.
Recap notes RBMS
● RQ → hypotheses → study design & data collections → descriptive statistics → inferential
statistics → conclusion → RQ
● What is null hypothesis testing? We create a H0 where there is no effect/association. Then we
test P(data|H0) and if that is high, then we keep H0 (note that we don’t accept it as true. We
simply retain it). If low, we reject H0.
● What is an alternative hypothesis? There is an effect/an association
● How does a well formulated RQ look? Population, Intervention, Comparison, and Outcome.
● What is a P-value? A P-value is the probability of getting the observed data assuming the null
hypothesis is true. Low P-value (P<0.05)=
reject H0. High P-value (P=>0.05) = retain
H0.
● What is a two-tailed vs a one-tailed
hypothesis? Two tail hypothesis → about
an ‘effect’ or ‘difference’ caused by A on
B. Could be an increase or a decrease -
bidirectional. One tail hypothesis → more
specific of the direction. E.g. “smaller
than”, “greater than”, etc.
● What are the two kinds of biomedical
research designs? Observational and
experimental
● What are the kinds of observational
experimental designs?
Cross sectional = observational in the
moment study
Cohort/prospective study = follow into
the future
Case control = observational comparison case and control
● What are the kinds of experimental studies?
Randomised control trial = randomly assorted to 1 of the 2 conditions
Cross-over study = participants are in both conditions but with a washout period.
● What is the goal of descriptive statistics? To present, organise and summarise the data
observed in the sample.
, ● What are the descriptive statistics used? Frequency (or proportion as a %), mean, median,
mode, variance, range, IQR, STDEV + graphs and figures
● What are inferential statistics for? To draw conclusions about a population based on data
observed in a sample
● What is used in inferential statistics? Statistical tests that results in a p-value = probability of
the data given that the H0 is true.
Very unlikely (below 0.05) → reject H0. That means low p-value means there is
significance.
Hence, alpha = 0.05 (5%)
Statistical tests
● What are the 2 broad groups for
statistical tests? Test of association
(are two variables related) and test of
difference (Do two or more groups
differ → comparing proportions or
means)
● What are the tests of association?
○ Chi squared test for
independence → tells you if
there is an association
between 2 nominal variables
(e.g. smoking yes/no, lung
cancer yes/no)
○ Pearson’s R: measures
strength of linear relationship.
Assumes normal distribution.
○ Spearman’s Rho: used for
association when Data are not
normally distributed and
Relationship is monotonic but
not perfectly linear + data is
ordinal
“Does tumor stage (I–IV) correlate with survival time?”
○ Regression analysis: models how one variable predicts another. “How much does
cholesterol increase risk, controlling for age?”
● What are the tests of difference?
○ Binomial test/One sample Z test: Compare sample proportion to population
A vaccine trial expects 95% immunity, but your sample shows 90%. Is 90%
significantly different from 95%?
○ Chi squared goodness of fit test: Compare observed frequencies to expected
frequencies
○ Two-sample Z test: Comparing proportions between two independent groups.
Comparing infection rates: Hospital A: 8%, Hospital B: 5%
Is Hospital A significantly worse?
, ○ Chi squared test homogeneity: Comparing distribution of a categorical outcome
across groups
Do treatment groups differ in
side-effect type?
○ One sample T test -
comparing sample mean to
known reference
○ Independent samples T test -
comparing means of two
independent samples
○ Dependent samples T test -
comparing means in same
sample before and after
● What are the features of a normal distribution? Mean 0 and STDEV 1.
● What do Z scores help with? Basically a standardised version of an observation. Z scores
helps to compare two scores from two different distributions. Z score is basically the SD
away from the mean of each situation
● How does a t-distribution work? Basically you
have a STDEV of the sample, which you
assume is equal to the population’s STDEV.
You calculate a t value. There are 2 t-critical thresholds (t-critical-0.025 and t-critical-0.975)
on the normal curve that your calculated t-value has 95% chance of falling within. The
remaining 5% is for below 0.025 or above 0.975 which is very unlikely. So if your t-value
falls somewhere there, it means it’s very unlikely to get your results given H0 is true, and so
