The flashcards below were created by user
Rx2013
on FreezingBlue Flashcards.

Sample
 must be representative of the entire population
 generizability

Random sampling
 evenly distributed patient characteristics
 equal chance of being in the study

Paired sample
each subject has a matching mate in the other group

independent sample
 subjects are not matched as pairs
 larger number of patients needed

Data types
 interval
 ratio
 ordinal
 nominal

Interval data
 continuous
 known, equal distance between each interval
 may contain negative numbers
 ex. temperature

Ratio data
 continuous
 known, equal distance between each interval
 nonarbitrary zero
 no negative numbers
 ex. age, wt, BG, BP, incidence rates of outcomes

Ordinal Data
 categorical
 ranked
 intervals not equal
 ex. pain scale, military rank

Nominal Data
 categorical
 cannot be ranked
 binomial or nonbinomial
 binomial ex. mortality, gender
 nonbinomial exp. eye color, hair color, ethnicity

statistical methods/test are described as being
parametric or nonparametric

When do you use parametric tests?
 data from sample results numerically describe a characteristic
 assumes normal distribution
 applies to most interval and ratio data

which are more powerful, parametric tests or nonparametric tests?
parametric tests

when do you use nonparametric tests?
 no assumption reguarding distribution
 applies to ordinal and nominal data
 can be applied to interval and ratio data when distribution is skewed

specific characteristic determining use of parametric/nonparametric tests
 type of data
 assumed distribution of data
 how many groups a study includes

parametric test examples
 student's ttest  2 groups
 anova, ancova  >2 groups

nonparametric test examples
 chi square test
 fisher's exact test  smaller #
 mannwhitney U test
 Wilcoxin test
 Kruskalwallis test

Null hypothesis
 theory about an outcome that a study is designed to test
 null = no difference between study groups

Alternate hypothesis
 opposite of null hypothesis
 there is a difference between study groups

alpha
level of significance

we usually classify something as statistically significant if it
falls below a certain level of significance

statistical & clinical significance
 may be statistically significant but not clinically important
 if it is not statistically significant it cannot be clinically important

pvalue
 chosen level of significance
 usually p<0.05
 describes the probability that differences between 2 groups occured by chance

if pvalue < alpha
 statistically significant
 reject null hypothesis

if pvalue > alpha
 not statistically significant
 fail to reject null hypothesis

setting alpha establishes
probability of making a type I error

type I error
 incorrectly reject null hypothesis
 false positive
 say there is a difference when there isn't

setting Beta establishes
 probablility of making a type II error
 conventionally set at 0.2

Type II error
 incorrectly fail to reject the null hypothesis
 false negative
 say there is no difference when there may be one
 associates with too small sample population
 used in calculating power

