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

Type of data in nonparametric tests

Nominal data
 mutually exclusive
 yes/no
 male/female

Chi squared test
 two samples
 at least 40 observations (20 per group)
 nominal data

Fisher's Exact
 similar to chi squared but for smaller sample sizes
 two groups
 less than 5 per group
 usually for uncommon events
 for nominal data

MantelHaenszel
 three or more groups
 series of chisquares
 totals are added up and overall chisquare is calculated
 for nominal data

McNemar's Test
 similar to chisquared
 data is paired (there is a control or pt serves as their own control)
 for nominal data

Ordinal Data
 categorical
 ranked on a scale with no consistent distance between values
 ex. pain scale

MannWhitney U/Wilcoxon Rank Sum
 two groups
 ranks individual observations rather than values
 data values are assigned from lowest to highest
 the sum of the data is totalled at the bottem fo the chart
 (rank pain 16 and the group with lower number has better pain control)

KruskalWallis
 for observational data
 for multiple comparisons
 if average ranks are close H_{0} is not rejected
 post hock analysis needed to determine difference between treatment groups
 Friedman's test is similar for 3 or more groups

Wilcoxon signed rank
 similar to paired ttest
 for paired data
 differences arranged in order of absolute value
 observations of 0 are excluded
 average rank is assigned

continuous data
 ratio or interval data
 placed on a numerical scale with infinite amount of values
 equal distance between those values

Ttest
 "student's Ttest"
 compares the mean of a population to the known mean
 2 groups
 variance of both groups must be equal
 interval or ratio data

ANOVA
 for interval or ratio data
 3 or more groups

ANCOVA
 for ordinal or ratio data
 similar to ANOVA but controls for covariates
 ex. study effect of bisphosphonates on osteoprosis, but age and exercise are also factors, control for these to see the difference the drug makes

Paired Ttest
 for interval or ratio data
 compare the mean difference between 2 groups that are paired or matched
 can be used when the subject is their own control (cross over)

ANOVA for repeated measures
 similar to ANOVA but measures subjects under a different set of conditions in a repeated manner
 ex. pt with osteoporosis screened anually for next 5 years for changes in BMD

correlation
 measure of association between 2 variables
 r ranges from 1 to 1
 independant variable is x
 dependent variable is y

regression
 0.91 = excellent correlation, probably predictive
 0.70.9 = good correlation, may be predictive
 0.60.7 = good correlation, may not be predictive
 <0.6 considered not predictive

Kaplan Meier survival analysis
 survival analysis as a function of time
 interval or ratio data
 time on the x axis

Survivial analysis censored vs. uncensored
censored means incomplete data, but is analagous to intenttotreat

PostHoc Analyses
 performed after primary analysis when H_{0} has been rejected
 considered inferior to planned comparisons
 often easier to understand

