Stats Flashcards
From all resources: AAP PREP:EM Course
formula for NNT
NNT = 1/ARR ARR = absolute risk reduction
type I error
rejecting the null hypothesis when it’s actually correct
false positive study
alpha
type II error
failure to reject the null hypothesis appropriately
false negative study
beta
failure to reject the null hypothesis appropriately
type II error
rejecting the null hypothesis when it’s actually correct
type I error
What 3 factors do you need to look at to determine the appropriate type of statistical test to perform?
- is the distribution parametric or not?
- is your data continuous, nominal/ordinal, or categorical?
- are your tested populations dependent or independent?
Sensitivity =
TP / TP+FN
Specificity =
TN / TN+FP
Power
1 - beta
beta = type II error rate
PPV =
TP / TP+FP
NPV =
TN / TN+FN
Incidence vs Prevalence
Incidence = RATE of new diseases over a period of time Prevalence = number of existing disease cases at a specific POINT in time
What do you need to calculate the sample size for a study?
- the effect size
- the type I error rate
- the type II error rate
what is the reciprocal of the rate difference?
the NNT
What is the other name for the Mann-Whitney U test?
Wilcoxon rank-sum test
What is another name for the Wilcoxon rank-sum test?
Mann-Whitney U test
What is the t-test?
statistical test for parametric, continuous data that is independent
what is the paired t-test?
statistical test for parametric, continuous data that is paired
what is the ANOVA?
statistical test for parametric, continuous data with 3 or more independent groups
what is the wilcoxon signed-rank sum test?
statistical test for non-parametric, continuous data that is paired
non-parametric corollary to paired t-test
what is the wilcoxon ranked-sum test?
statistical test for non-parametric, continuous data that is independent
nonparametric corollary to t-test
also called the mann-whitney u test
what is the kruskal-wallis?
statistical test for non-parametric, continuous data with 3 or more independent groups
nonparametric corollary to ANOVA
what is the Chi-square test?
statistical test for parametric, categorical data with independent groups
what is the Fisher’s exact test?
statistical test for parametric, categorical data with independent groups if there are <5 measurements/group
what is logistic regression?
statistical test to predict the relationship between a DICHOTOMOUS outcome vs a set of variables while controlling for other variables in the analysis
what is Bonferroni’s correction
a method to correct for multiple repeated testing on the same data set
what is Kolmogorov-Smirnov?
test to determine if data is parametric or not
what is Shaprio-Wilk?
test to determine if the data is parametric or not
LR +
= (positive test/presence of disease) / (positive test/absence of disease)
= sensitivity / (1 - specificity)
LR -
= (negative test/presence of disease) / (negative test/absence of disease)
= (1 - sensitivity) / specificity
How do you use LRs?
You have to convert the pretest probability to odds
This is pretest probability / (1 - pretest probability)
Then you multiply by the LR
Finally, you convert the odds BACK to the probability!
What are clinically significant LRs?
LR + > 10
LR - < 0.1
formula for odds ratio
(AxD)/(BxC) in standard 2x2 table
what’s the difference between odds ratio and relative risk?
relative risk is used when patients are followed over time.
odds ratio is used when patients already have the outcome and you look back retrospectively at an exposure of interest