Exam 2 Flashcards
Descriptive stats purpose
Summarize or describe the main features of a dataset
(Central tendency, measures of spread, data visualization)
Inferential stats purpose
Makes inferences and predictions about a population based on a sample
(Hypothesis testing, confidence intervals, p values)
Random sampling
All members of the population have an equal chance of being selected
Cluster sampling
Selecting random samples from certain natural groupings
Systematic sampling
Selecting every random individual from a list
Convenience sampling
Selecting a population based on convenience of the investigator
Hypothesis testing
Used to make inferences or decisions about a population based on sample data
Allows to test assumptions and determine if there is enough evidence to support a claim
Null hypothesis testing (H0)
States there is no stat difference between groups
Goal to disprove or reject the null hypothesis
Test this first then test alternative
Alternative hypothesis testing (H1)
Suggests there is a stat difference between groups
Hope to prove or accept alternative hypothesis
What is a type 1 error
False positive
Occur when null hypothesis is incorrectly rejected
Denoted by alpha**
(When alpha = 0.05 then 5% chance of making a type 1 error)
What is a type 2 error
False negative
Occurs when null hypothesis is incorrectly accepted
Means that no difference when there is one
Donated by beta
(Risk of type 2 error increases with a smaller sample size)
If p < alpha the null is what
Rejected (significant result)
If p > alpha the null is what
Null is not rejected (not significant)
What is power calculation
The probability of correctly rejecting the null hypothesis
Power = 1 - Beta
Higher power reduces type 2 error, increase sample size boosts study power
If beta is 0.2 then the study power is what? If beta is 0.1 the power is what
80%
90%
Stat significance is or is not equal to clinical significance
Is NOT
Confidence intervals (CI)
Range of numbers likely to contain the population interval
A method to determine stat significance: studies may report CI, p-value, or both
(Usually reported as 95% confidence interval)
If evaluating differences between groups the CI cannot cross what? If evaluating proportions, RR or OR, the CI cannot cross what?
0
1
Is normally distributed parametric or non-parametric
Parametric
(Non normally distributed is non parametric)
What is students t-test
Parametric test
Have to be normally distributed and continuous data
Random samples from two independent populations
Compares means***
What is a paired t-test
Determine differences between the same individuals (dependent or paired)
Dependent variable is normally distributed and continuous
Compares means
What is analysis of variance (ANOVA)
3 or more independent samples
Continuous parametric data
What is Mann Whitney U (Wilcoxon Rank Sum test)
Two independent samples
Ordinal and continuous non parametric data
Wilcoxon signed ranks
Two related paired samples
Ordinal and continuous non parametric data
Kruskal Wallis
3 or more independent samples
Ordinal and continuous non parametric data
Chi squared test
Two independent samples or 3 or more independent samples
Nominal data
Fishers exact test
Two independent samples
nominal data
McNemar
Two related paired samples
Nominal data
Subgroup analysis
Explores treatment across different subgroups of a population
Goal to see if effect of treatment is consistent across all groups
Univariate analyses
Comparison of one dependent variable to one independent variable
-doesn’t account for confounders
-may have stat significant results that are not accurate
Multi variable analysis
Compare one dependent variable to multiple independent variables
-multiple variable = confounders or covariants
-able to determine stat significance
Multi variable linear regression
Dependent variable is a continuous variable
Independent variables can be nominal, ordinal, or continuous
P values will be provided for each covariant and the relationship to the dependent variable
What is multicollinearity
Correlations exist between the variables indicating redundancy and potential bias (relationship exists between variables)
Multi variable logistic regression
Dependent variable is dichotomous variable
Independent variable can be nominal, ordinal, or continuous
Relationship is non linear
Survival analysis
Methods to determine time to an event
If event is not experienced subject is surviving
If event is experienced subject is failing
Dependent variable is time
Censoring meaning
Done for incomplete data
-If subjects are lost to follow up or do not have event they are censored
-data for that patient included up until end of study or lost to follow up
Kaplan Meier method
Survival analysis
-can’t account for relationship of covariates on survival
-analysis cannot determine if the independent variable predicts survival
Kaplan Meier method
Estimation of survival curve
What are the two ways to present the Kaplan Meier method
Survival function: cumulative frequency of subjects not experiencing event
Hazard function: cumulative frequency of subjects experiencing the event
Cox proportional hazards model
Stat comparison accounting for covariates
Log rank test
Stat comparison
Matching test (Propensity score matching)
cases matches to controls based on investigator determined factors
help remove differences in baseline characteristics and only evaluate study question
IRB definition
Committee ensuring all study subjects are protected and not exposed to unnecessary harm and/or unethical medical procedures
True or False: all research projects that use humans as study subjects must be IRB approved before beginning study enrollment
True
Aka ethics committee
At time of study enrollment, a study subject needs to be given an informed consent form that informs them of what 6 things
Study procedures
Rights and responsibilities for study participation
Risk/benefit of study population
Compensation for study participation
Participation in the study is voluntary
Right to withdraw from the study
Cohort studies: subject grouping is done via what
Exposure
(Different from case control studies)
What is the importance of a cohort studies follow up
Subjects followed to determine the incidence of an outcome (symptom, disease, death)
Relative risk
The risk of developing a disease or event in subjects exposed to a variable versus those not exposed to
Prospective cohort study (aka concurrent)
Subjects identified based on current OR past exposure and then followed for a period of time to determine if subject had the predetermined outcome
Retrospective study (aka non concurrent)
Exposure and outcome have already occurred
Data obtained from a database to determine patient exposure and outcomes
Ambidrectional cohort study
RARE cohort study design
Subjects identified and outcomes assessed in the past and future
Confounders
Variable (factor) that influences or is related to other variables in the study
How to fix no randomization in cohort studies
Limit inclusion/exclusion criteria
Prospective studies in cohort studies are subject to ______ from more intensive monitoring
Bias
Case control studies: subject grouping is done by what
Outcome
(Subjects with outcome are compared to similar subjects without the outcome)
Case control studies follow up importance
Data collected from subjects with and without outcome to determine if they experiences the exposure
(Via patient interview or survey OR by reviewing data records)
Both exposure and outcome known at start of study
Odds ratio
Estimate of the relative risk
(Done in case control studies evaluation)
Retrospective case control study
Exposure and outcome has already occurred
Data obtained from a database or pt interview to determine patient exposure and outcomes
How to use case control studies in practice
Identify risk factors of a disease, benefits and safety risks of both med and procedure
For rare disease
Less expensive
Is randomization possible in case control studies
No
How to fix case control studies randomization issues
Matching
(Cases matches to controlled based on investigator determined factors)
Outcome measurements may not be accurate in case control studies what can you do with study investigators to confound for this
Blind them to subjects exposure status when determining outcome
One technique to overcome historical data not being as accurate is to assign multiple controls to one case patient to help minimize bias, what study is this in
Case control
In cross sectional studies data is only collected at ____ time point for all subjects
ONE
How is evaluation determined in cross sectional studies
Descriptive stats (mean or median)
How to use cross sectional studies in practice
Useful for measuring current health status or disease control
What study design is good for hypothesis generating
Cross sectional
Inclusion criteria similarities between RCT and observational studies
Subject demographics or characteristics that MUST be present in order to be enrolled in study
Impact on study findings similarities between RCT and observational studies
Inclusion and exclusion criteria impact study validity
What is active control
Standard of care
Composite endpoints
Group of endpoints that are considered clinically important
Primary or secondary endpoints
Measure overall effect of therapy
Overstate results of study
Surrogate endpoint
An endpoint that is associated with an outcome but does not directly measure it
Subgroup analysis
Dividing the study population into subsets of participants based on common characteristics
What study is single case, 3 or fewer cases, more than 3 cases
single case: case report
3 or fewer cases: case series
more than 3 cases: cohort study
What studies are not under the jurisdiction of IRB
Case studies and case series
Importance of case reports
Generation of new hypothesis
Improvement of medical education
Early benefits or harm that might take additional time if reviewed in a more formal setting
Medical writing and editing skills
What study describes an account of an individual clinical case that is accosted with an uncommon presentation, clinical course, or outcome
Case report
What study should allow the reader to critique and reach conclusions about the case prior to reading the discussion section of the manuscript
Case report
What are the contents of a case report
Title
Background
Patient case
Discussion
Conclusion
For case studies do you have to write a report for each patient
Yes need to write a report for every person you do, you cannot add them all in one report