Stats Flashcards
Data that is ordered or ranked categories (stages, functional status, etc)?
Ordinal
Data that is categorical (male/female, alive/dead, etc)?
Nominal
Data that is continuous (equal distance between values)?
Interval
Central tendency for interval (continuous) data?
Mean (average)
Central tendency for ordinal data?
Median (1/2 subjects above and 1/2 below)
Central tendency for nominal data?
Mode (most frequently occurring category)
Parametric or Non-parametric?
Continuous Data
Normal distribution or n>30
Parametric Statistics
normal distribution
Parametric or Non-parametric?
Ordinal or Discrete Data or n<30
Tests have less power
Non-parametric Data
Choose a Test
2 groups Continuous (parametric) data
Student’s t-test
Choose a Test
2 groups Nonparametric data
Mann-Whitney U or
Wilcoxon rank sum
Choose a Test
2 groups Categorical (Nominal)
Chi-square
Fisher’s Exact Test
Choose a Test
3+ groups Continuous (parametric) data
One-way ANOVA
Choose a Test
3+ groups Nonparametric data
Kruskal Wallis
Choose a Test
3+ groups Categorical (Nominal) data
Chi-square test
Choose a Test
Predict a variable from another with continuous (parametric) data
Linear Regression
Choose a Test
Predict a variable from another with categorical (nominal) data
Logistic Regression
Choose a Test
2 paired groups - continuous (parametric) data
Paired t-test
Choose a Test
2 paired groups - nonparametric data
Wilcoxon signed rank
Choose a Test
2 paired groups - categorical (nominal) data
McNemar’s
Choose a Test
Association btwn 2 variables - continuous (parametric) data
Pearson correlation
Range -1 to 1 0=no corr
Choose a Test
Association btwn 2 variables - nonparametric data
Spearman correlation
Range -1 to 1 0=no corr
Choose a Test
Association btwn 2 variables - categorical (nominal) data
Chi-square
This type of analysis analyzes “time to event” (death, recurrence, etc)?
Survival Analysis
1group - Kaplan Meier Survival Curve
2groups - Log-rank test
Predict variable from another -
Cox proportional hazards regression
(Hazard ratios)
What measures the chance that the difference in a study occurred by chance?
p Value
What measures the probability that the difference between two values could have occurred by chance?
(Type 1 error - false-positive conclusion)
p Value
Risk of disease in exposed
———————————— = ?
Risk of disease in unexposed
Relative risk
Cohort studies and RCTs
Odds of disease in exposed
———————————— = ?
Odds of disease in unexposed
Odds ratio
Will overestimate RR with common events
With regard to relative risk and confidence interval, a value of 1 indicates?
No effect
> 1 - increased risk/odds
<0.05
What is a Type 1 error?
False-positive: detect a difference when no exists
Probability of incorrectly rejecting the null hypothesis
What is a Type II or beta error?
False-negative.
Probability of not detecting a significant difference when one exists
Power is?
1-beta
Probability of detecting a difference when one exists
What do you need to calculate sample size?
1 frequency of X in group A&B
2 alpha (ie 0.05)
3 power (ie 0.8 or 1-beta)
4 ratio of A to B
Can calculate any one of 5 from the other 4
Which varies based on prevalence of a disease, sensitivity/specificity or positive/negative predictive values?
Predictive values
Sensitivity and Specificity are properties of a test and not affected by prevalence
Sensitivity is?
Positive test in disease
True positive test over everyone with disease
Specificity is?
Negative test in health
True negative test over everyone with a negative test
Answers question - if a test is positive then what percent of people will have the disease?
Positive predictive value
Answers question - if a test is negative then what percent of people will not have the disease?
Negative predictive value
Positive predictive value is?
True positive test over all positive tests
Negative predictive value is?
True negative test over all negative tests (true and false)
