Multivariate Statistical Analysis and Modeling Flashcards
what are statistics used for?
to discern justifiable differences in quantitative patterns
give 3 classic statistical tests and why they are not practical for epidemiology
- basic t tests
- ANOVA
- Chi squared
too burdensome to apply to situations where multiple variables and their interactions are in question
describe the T test and give an example of how it could be used
tests whether significant differences, not attributable to random chance, are present in continuous data; could use for averaging hemoglobin levels in two different groups of people
what does the chi squared test do? give an example
tests significance between groups of non-continuous data
examples are infected versus not infected, vaccinated or not vaccinated
what are chi squared tests useful for?
2x2 contigency tables
what is Fisher’s exact test?
tests for significance in much the same way as the chi squared test does, but works with ANY SIZE DATA SET!!
what does the Fisher’s exact test give?
gives an exact P value rather than an approximation like other tests do
what can Fisher’s exact test also be applied to?
2x contigency tables
what is mutlivariable regression? (2)
- multiple independent variables are regressed on a single dependent variable
- all variables are considered together to discern correlations either individually or in combinations
what does multivariate analysis allow for?
allows more than two variables to be analyzed at once
is multivariable regression a type of multivariable analysis?
not usually considered under the heading, but is technically multivariate analysis
what are the 2 general types of multivariate analysis technique?
- asnalysis of dependence
- analysis of independence
describe the analysis of dependence multivariate analysis technique
where one or more variables are dependent variables, to be explained to predicted by others
give 3 analysis of dependence multivariate analysis tests
- multivariable regression
- PLS (partial least squares)
- MDA (multiple discriminant analysis)
where is multivariable regression analysis often seen?
in cohort studies
describe the analysis of independence multivariate analysis method
no variables are thought of as dependent, just look at the relationship among variables, objects, or causes
give 2 analysis of independence multivariate analysis tests
- cluster analysis
- factor analysis
in most cases, why is data anlaysis done?
to test the association between two or more variables (bivariate)
what is often the aim of analysis in health research?
- to establish the association between risk factor and disease
- to establish the association between therapy and outcome (patient oriented research)
what are the 2 types of regression models?
- simple linear (y = mx+b)
- multiple linear regression (y = b + m1x1 + m2x2…)
what is a model?
a set of theoretically and/or evidence based associations between variables
what does data analysis test?
data analysis tests the extent to which the proposed theoretical associations are observed in the data
what is model development based on?
model development is based on previous knowledge (previous research evidence about associations)
what is model development mostly about?
model development is mostly about deciding which independent variables (risk factors) should be in the multivariate analysis
what do SIR models mean/stand for?
S: susceptible, S(t): the disease-susceptible people in a population
I: infectious, I(t): those capable of transmitting an infection to others
R: recovered R(t): those that have been infected and recovered
what do vaccines do in terms of the SIR model?
try to move people from susceptible to recovered (since vaccine is like a mini infection)
what are the 3 assumptions of the Kernack-McKendrick model?
- assumes a well-defined, closed, and fixed population
- assumes an infected person is instantaneously infective and infectivity lasts duration of disease state
- assumes a homogenous population with no genetic variation or demographic differences