Advanced Biostatistics 1 Flashcards
What type of data?
Nominal
Ordinal
Continuous
What type of distribution?
Normal
Poisson
Binomial
Define dependent variable
The variable you are most interested in
The outcome
Define independent variable
The variable that affects the dependent variable
Covariates
Predictors
Risk Factors
Define Nominal
Data divided into mutually exclusive groups
Yes/No
Female/Male
Define Ordinal
Prescence of clearly defined hierarchy
Military rank
Define continuous
Interval Scale (temp) Ratio (mass, weight, pressure)
Define Parametric
Data can be described with 2 parameters (mean and SD)
Statistical test
What do statistical test do?
Compare the equivalence of parametric definers More powerful Many assumptions (normally distributed, homoscedastic groups)
Non-parametric
Compares equivalences of ranks
1 assumption: independently and randomly collected samples
Require larger sample sizes to reach power
Difficult to manipulate if there are ties
Define Independent sample
Subjects are chosen independently from a population
Define Paired Sample
Units in teh samples are carefully matched in pairs so that each pair are as similar as possible (twin studies)
Statistical Test for Nominal data?
X2
Fisher’s exact test (ns (paired)
Mantel-Haenszel (controls confounders)
Statistical Test for Ordinal data?
Vilcoxon rank sum (Unpaired 2 grps) Mann Whitney U (2 grps) Wilcoxon signed rank (paired) Kruskal Wallis (3+ grps) Friedman (crossover)
All work on continuous data except friedmans
Statistical Test for Continuous data?
Student’s t-test (2 grps or corrected by Bonferoni’s)
Analysis of variance (2+ grps)
Define correlation
Examines the linear strength and direction of the ASSOCIATION between two variables
r and r^2
Pearson product moment correlation coefficient
r
Normally distributed
-1 to 1
Coefficient of determination
Determines the amount of variation
0-1
Define regression
Examines the ability of one or more variables to predict another variable (independent associations in non-randomized studies)
Univariate or multivariate
Univariate regression
risk of MI in chest pain, rist of GI bleeding in critically ill
Multivariate regression
Modeling many possible predictive variables to ascertain which ones predict target variable
Adjust for confounders
Linear Regression
Relationship between a continuous outcome and predictors
y=mx=b
Logistic regression
Predicting a categorical outcome from one or multiple predictors
logy= mx+b
Define survival analysis
Time between study entry and an event
Takes drop outs and lost to f/u into account
Kaplan-Meier curve
Log-Rank test and Cox proportional hazards
