Stats Flashcards
Absolute Risk
of events in treated or control groups
/
# of people total in that group
Absolute Risk Reduction
absolute risk in control group (ARC) - absolute risk in treatment group (ART)
Group A
-Have disease
-Have exposure
Group B
-No disease
-Have exposure
Group C
-Have disease
-No exposure
Group D
-No disease
-No exposure
Relative Risk
ART
/
ARC
OR
A/(A+B)
/
C/(C+D)
Relative Risk Reduction
1 - RR (x 100%)
Odds Ratio
A/C
/
B/D
Odds Ratio Interpretation
< 1 = negative association (protective)
1 = no association
> 1 = positive association (harmful)
Number Needed to Treat
1/ARR
Sensitivity
ability of a test to detect a true positive
TP/(TP+FN)
Specificity
ability of a test to detect a true negative
TN/(TN+FP)
Positive Predictive Value
proportion of those w/ disease when test is positive
TP/(TP+FP)
Negative Predictive Value
proportion of those w/o disease when test is negative
TN/(TN+FN)
Mortality Rate
deaths during specified time interval
/
population size at risk of death
Incidence Proportion
new reported cases
/
initial population at risk
*expressed as %
Incidence Rate
new reported cases
/
susceptible population
*must subtract those who already had disease from susceptible population
Type I Error
concluding there is a difference when one does not exist (rejecting null hypothesis)
probability of making this = alpha
Type II Error
concluding there is no difference when one actually exists (null hypothesis not rejected when it should be)
probability of making this = beta
Continuous Data
Interval - ranked in specific order + arbitrary zero (ex: degrees F)
Ratio - ranked in specific order + absolute zero (degrees K, weight, BP)
Categorical Data
Nominal - classified into groups in unordered manner (ex: race, sex, presence/absence of disease)
Ordinal - ranked in specific order (ex: NYHA class, pain/Likert scale)
Statistical tests used for nominal data
2 Groups (Independent) = x2 or Fisher exact test
3+ Groups (Independent) = x2 or Fisher exact test
2 Groups (Dependent) = McNemar test
3+ Groups (Dependent) = Cochran Q test or Mantel-Haenszel test
Chi-Fish, Chi-Fish, McNem, Cock-Mantel
Statistical tests used for ordinal data or continuous data (not normally distributed)
2 Groups (Independent) = Mann-Whitney U test or Wilcoxan rank sum test
3+ Groups (Independent) = Kruskal-Wallis test
2 Groups (Dependent) = Wilcoxon signed-rank test
3+ Groups (Dependent) = ANOVA
Man-Wit Cox Rank, KW, Cox Sum, ANOVA
Statistical tests used for continuous data (normally distributed)
2 Groups (Independent) = Student t-test
3+ Groups (Independent) = ANOVA test
2 Groups (Dependent) = Paired t-test
3+ Groups (Dependent) = ANOVA
When to use mean
continuous + normally distributed data
When to use standard deviation
continuous, parametric data
When to use median
ordinal or continuous data
When to use mode
nominal, ordinal, or continuous data
