Epidemiology and Biostates Pt. 4 Flashcards
Regression analysis
used to estimate how the value of one variable corresponds to a value of another variable. A simple linear regression model uses X as the fixed non-random independent variable and Y dependent variables which are normally distributed. ie. predicting milk yield (Y) by parity in dairy cows (X).
Multiple linear regression model
models the linear relationship between multiple independent variables (X’s) and a dependent variable (Y)
Logistic Regression Model
used to model the relationship between an independent (quantitative or qualitative) variable and a dependent outcome variable that is discreet, often dichotomous. The outcome is expressed as an odds ratio
Fisher’s Exact Test
used as a test of homogeneity, but when sample sizes are small since Chi-square tests are not accurate under small sample sizes
Chi-Square Test of Association
done to determine whether or not there is an association between 2 or more proportions in differing populations (ie. test if there is a difference in color blindness between males and females)
Chi-Square Test of Homogeneity
method for testing the null hypothesis that 2 or more population proportions are equal
when are survival analysis studies done?
to monitor the progress or survival of patients undergoing a surgery or tx regimen. Can be used to compare median survival times for patients receiving competing medications
Kaplan-Meier Procedure
determines the probability of surviving for a specific length of time considering censored data from the patients in the study
Sign Test
considers the positive and negative outcomes of a trial or study and converts these +/- counts to a Binomial distribution. The probability of the number of positive or neg outcomes is then compared to the null hypothesis that the median of the differences should be zero
Wilcoxon Signed Rank Test
can be used in the same setting as the sign test but it also incorporates the differences from the median in each data point, making it more informative and giving it more power than the sign test
Type I error
rejecting the null hypothesis when it is true
probability of Type I error = a = level of significance
Type II error
accepting the null hypothesis when it is false
probability of Type II error = B (beta)
Power
probability of rejecting the null hypothesis when it is false (1-beta)
-a measure of the ability of a study to detect a true difference
-should be calculated prior to carrying out a study to determine the number of observations needed to detect a desired degree of difference
T/F: if a study has an inadequate sample size, then the result with a null finding is uninformative
T
case fatality risk (aka case fatality rate)
the proportion of persons with a particular condition (case) who die from that condition.
-A measure of the severity of the condition
-numerator is restricted to deaths among people included in the denominator
= # cause-specific deaths among the incident cases/number of incident cases
point source epidemic curve
epi curve with a sharp upward slope and gradual downward slope. Usually the exposure period is relatively brief and all cases occur within one incubation period
propagated epidemic curve
usually has a series of progressively taller peaks, each an incubation period apart. Happens in outbreaks that are spread from person to person and that may last longer than common source epidemics
types of random sampling
1) stratified - layering a study population by each variable (ie. sex, age) we consider important and then randomize participants to tx groups within each stratum
2) simple - a subset of a population in which each member of the subset has an equal probability of being chosen. A simple random sample is meant to be an unbiased representation of a group
Pros/cons of stratified sampling
Pros:
-provides greater precision than a random sample of the same size and often requires a smaller sample size
-prevents an unrepresentative sample
-ensure we obtain sufficient sample points to support a separate analysis of any subgroup
Cons:
-more admin effort
-analysis is computationally more complex
Pros/cons of simple sampling
Pros:
-more likely to represent the whole population
-quicker, cheaper, easier
-involves less judgment
Cons:
-risk of selecting samples from a few variations only
-redundant and monotony
Steps of risk analysis
1) risk assessment (including determining whether a certain chemical is casually linked to particular health effects, performing a dose-response assessment, determining the extent of human exposure before or after the application of regulatory controls, and describing the nature and magnitude of the risk)
2) risk characterization (the estimated incidence of the adverse effect in a given population)
3) risk management (involves the evaluation of alternative regulatory actions and the selection of strategy to be applied)
4) risk communication
What is bias?
an error in methodology that results in a systematic deviation of the result away from the truth
random error is aka variability, random variation, or “noise in the system”
-can be minimized with large sample sizes
-has no preferred direction, will average out to having no net effect
Internal vs. external validity
internal: ability of a study to measure what it intends to measure
-must be present to extrapolate to a different population
external: ability of a study to be extrapolated to a target population
-impacted by selection bias
types of bias
confounding
information bias
selection bias
admission bias
prevalence/incidence
volunteer (self-selection)
losses to follow-up
inappropriate comparison group
confounding
a mixing of effects that suggests an association where none exists or masks a true association
-occurs when a variable is associated with the exposure of interest and the outcome of interest, but is not the causal pathway
information bias
results from problems with the info collected in the study, such as faults in the questionnaire
selection bias
a problem with who is in the study, such as inappropriate detection of participants, self selection, non-response bias
methods to control confounding
1) restriction/exclusion (selecting study participants with only one level of the confounding variable to eliminate effects of that variable)
2) matching (equalizes the frequency of the compounding variable in the groups being compared; select controls that are identical to cases in certain important characteristics)
3) analysis (collect data on potential confounding variables while conducting study-study data and then stratify the data based on these confounding variables)
Reliability
a measure of the repeatability or reproducibility of a clinical measurement. Aka precision