Second midterm content Flashcards
descriptive statistics vs inferential statistics
descriptive: used to describe the main features of a data set
inferential: used to draw conclusions and involve making inferences about the larger group from a smaller subset
measures of tendency vs measures of dispersion
tendency: mean, median and mode
dispersion: range. interquartile range and standard deviation
what is the central limit theorem
sampled means will tend to be normally distributed given that we took enough samples
- even If the data itself id not normal the CLT tells us that the distribution of the sampled means will be
what is a P value
the probability of obtaining the observed results assuming the null hypothesis is true
- if the p-value is less than the alpha value, the result is statistically significant
what is a T-test
- compares the means of 2 groups
- used when you have a continuous variable
what is an ANOVA test
- compares the means of 3 or more groups
- used when you have a continuous variable
- won’t tell you which groups are different but will tell any of the groups are significantly different
what is a chi-squared test
- determines if there is a relationship between 2 categorical variables
- use when you are dealing with types of groups and want to know is one category is related to another
what is a Pearson correlation test
- assesses the association between 2 continuous variables
- use wan you want to determine the direction and strength of a relationship between 2 continuous variables
what is regression
shows how much the dependent variable will change when the independent variable is changed
types of regression
linear regression: models outcomes of continuous variables
polynomial regression: measures continuous variables with a non-linear relationship
logistic regression: model binary outcomes (either yes or no)
what is multiple linear regression
when the outcome is dependent on more than one variable
what do the measures of association and measures of effect measure
MoA: the ratio of 2 measures of disease frequency - null value is 1
MoE: the difference between 2 measures of disease frequency - null value is 0
what are the measures of association
risk ratio
odds ratio
incidence rate ratio
what are the measure of effect
risk difference
attributable proportion (exposed)
population attributable risk
population attributable fraction
what is risk ratio
compares the risk of getting disease in the exposed and not exposed group
what is odds ratio
measures the odds that an outcome will occur if exposed, compared to the odds of the outcome occurring if not exposed
confidence interval for measures of association
if the confidence interval includes 1 the association is not statistically significant
what is incidence rate ratio
the ratio of disease incident rate in the exposed group vs in the non-exposed group
what are person-time years
allows you to account for the individual time that each participant contributes
interpreting MoA
range = 0 to infinity
MoA = 1 means risk/odds/rate in exposed is equal to non exposed
MoA > 1 means greater in exposed individuals
MoA < 1 means less in exposed individuals
what is the case for risk ratio and odds ratio in rare diseases
RR ~ OR because a and c values are so small
what are measures of effect useful for
knowing how much disease could be eliminated by removing the exposure
what are measures of association useful for
investigating causation
what is baseline risk
disease in the non-exposed group OR amount of disease present not due to the exposure
what do measures of effect tell us about exposed groups
how mush od the total risk of disease is actually due to the exposure of interest
what is risk/rate difference
the difference in risk or rate of the outcome in the exposed and non-exposed groups
interpretation of risk/rate difference
the excess risk of getting disease due to exposure is…
what is excess risk
extra risk beyond the baseline risk
what is attributable proportion
the proportion of risk/rate of the outcome in the exposed group that is due to the exposure
interpretation of attributable proportion
…% of disease in the exposure group is due to the exposure
measures of effects in exposed groups vs in population
exposed groups: risk difference and attributable proportion
populations: population attributable risk, population attributable fraction
what is population attributable risk
the amount of disease in the entire population that is due to the exposure
prevalence of disease in population - baseline level of disease
interpreting MoE
range = -infinity to infinity
MoE = 0: exposure has no effect
MoE > 0: exposure positively associated with disease
MoE < 0: exposure negatively associated with disease
- null value is 0 because it is a difference not a ratio
population attributable fraction
proportion of disease in the entire population that is due to the exposure
PAR/prevalence of disease in population
interpreting population attributable fraction
…% of disease in the population could be eliminated if the exposure was removed
interpreting population attributable risk
the excess risk of disease in the population that is due to the exposure is…%
inductive vs deductive research
inductive: observations
deductive: intervening
what is quantitative research
- objective
- one truth
- observer doesn’t influence findings
- facts separated from values
what is qualitative reasearch
- subjective
- truth is subjective perception of findings
- findings are mutually created between observer and subject
- value lead
types of qualitative research methods
interview
focus group
ethnographic research
case study research
record keeping
qualitative observation
types of descriptive studies
case reports
case series
surveys
analytical vs descriptive studies
descriptive
- provide information about occurrence of disease without investigating associations
- generate hypothesis
analytical
- establish relationships between risk factors and the occurrence of disease, make comparisons between groups
- hypothesis testing
types of analytical studies
observational studies
experimental studies
what is an observational study
researcher has NO control over subjects being compared
exposed subjects already exist
what is an experimental study
researcher randomly allocates subjects to the groups being compared and intervened (adds exposure)
when would you choose an observational study over an experimental study
if exposure of interest is harmful
if it is expensive to administer exposure
if the exposure is hard to control
types of observational studies
cross-sectional
cohort
case-control
what is a cross-sectional study
- select participants without regard to exposure or outcome status
- then measure the exposure and outcome at the same time (a snapshot)
objective of cross-sectional studies
to estimate prevalence - how many subjects have exposure and how many have outcome?
prevalence vs incidence
prevalence: new and old cases (have the outcome)
incidence: new cases (get the outcome)
why can we only measure prevalence and not incidence in cross-sectional studies
- because we cannot establish temporality since we determine E and O at the same time
- exception = longitudinal studies
what are longitudinal studies (cross-sectional)
the same individuals are measured more than once, allows to establish incidence
advantages of cross-sectional studies
- determine prevalence of E and O
- study multiple exposures and multiple outcomes
- good for studying permanent factors
- no loss-to-follow up selection bias
- less potential for bias than case-control studies
- can measure ALL MoA and MoE EXCEPT for rate
disadvantages of cross-sectional studies
- not good for rare E, rare O or short duration diseases
- doesn’t measure incidence, ant establish temporal sequence
- selection, information and confounding bias
what is a cohort study
select participants based on exposure status
what is a cohort
a group of subjects that has a defined characteristic in common - exposure status
general design of cohort studies
- choose a group of disease free individuals
- classify them based on exposure
- follow over a time period
- compare the development of disease in E+ and E- group
retrospective vs prospective cohort studies
retrospective: establish from past records and follow-up to present day or more past records
proscpective: start from present day and follow into the future
Analysis in cohort studies
- measures incidence (new cases)
- can use ALL MoE and MoA
advantages of cohort studies
- good for rare exposures
- study several outcomes
- temporal sequence established in prospective studies
- provides incidence data
- no recall-bias with prospective studies
disadvantages of cohort studies
- bad for rare outcomes
- can study only one or a few exposures
- selection, information and confounding bias
- expensive and time consuming
- can have long follow up time: loss to follow-up is a major concern and non-response bias
addressing confounding bias in cohort studies
exclusion/restriction: subjects have same level of major CFV
matching: E- have same CFVs as E+
analytical citron: control for CFVs with stratification or multivariable regression
ways to ensure validity in in cohort studies
- selection of E+ and E- groups
- follow-up of cohorts
- objective diagnosis of disease status
how can we have more than 2 exposure groups in cohort studies
can have multiple E- groups or E+ groups with different levels of exposure
what is the biggest challenge in cohort studies
follow-up
what is a case control study
choose participants based on their disease status
what are cases and what are controls
cases: have the disease/outcome of interest
controls: do not have the disease/outcome of interest
- selection occurs at this level
- determine who was exposed and who wasn’t in both groups
why are case-control studies always retrospective
cases have already occurred do we must trace back if they were exposed or not
analysis of case-control studies
compare proportion of exposure in the cases with that in the controls
- CANNOT calculate incidence/risk therefore NO RR, but can still calculate OR
advantages of case-control studies
- fast and inexpensive (because retrospective)
- best for investigating source of an outbreak
- good for rare diseases/outcomes
- good for multiple risk factors for a single disease
disadvantages of case-control studies
- CAN’T calculate disease incidence or prevalence
- can only study one outcome
- study design most prone to all biases
- bad for rare exposure
- temporal relationship is an issue if using prevalent cases
what is a study base (case-control studies)
the population from which cases (and controls) are obtained
primary study base: the population from which the cases arise can be easily defined
secondary study base: one or more steps removed from the primary population
control selection from a secondary study base
- controls should be randomly selected from non-cases in the registry
- those controls should be selected from diagnostic categories not associated with the exposure of interest
use of multiple controls in a case-control study
controls come from different sourced e.g. hospital and neighbourhood controls
why can’t we calculate risk/rate ratio in case-control studies
because we set the risk of our disease by selecting participants based on their disease status
- must use OR to estimate RR because OR ~RR for rare diseases
factors for evaluating studies (critical appraisal)
Methodological: study design, statistical analysis, addressing bias
Non-methodological: relevance, originality, conflicts or interest
how to estimate PAF and AP in case-control studies
AP = (OR-1)/OR
PAF
Recall bias in case-control studies
- big problem if accuracy differs between cases and controls
- retrospective so need to ask people about their past
how to ensure validity in case-control studies
- proper selection of cases and controls, to boost power can pick 4:1 controls:cases
- ensure comparability
- unbiased history of exposure to the factor of interest
- control for confounding
what is critical appraisal
a systematic process used to identify the strengths and weaknesses of a research article in order to assess the usefulness and validity of research findings
- lacks a “gold standard”
questions asked in critical appraisal (10)
relevance?
adds something new?
what is the research question?
appropriate study design?
addresses bias?
performed as per methods stated?
hypothesis stated?
appropriate stats used?
justified conclusions?
conflicts of interest?
what is the difference between the 3 types of observational studies
the way that participants are selected