Exam 3 - Biostatistical Tests Flashcards
A research perspective which states there will be no true difference between the groups being compared.
Null hypothesis
Three attributes that differ between nominal/ordinal/interval/ratio data.
Order/magnitude, spacing/scale, rational absolute zero
Does nominal data have magnitude? Does nominal data have a consistent scale?
No; No
Does ordinal data have magnitude? Does ordinal data have a consistent scale?
Yes (i.e. it can be ordered); No (example = satisfaction survey)
Does interval/ratio data have magnitude? Does interval/ratio data have a consistent scale?
Yes; Yes (Tigger/Buzz = Units)
Black
Nominal
Red
Ordinal
Orange
Interval
Yellow
Ratio
True/False: After data collection you can move up (i.e. nominal => ordinal => interval => ratio) is specificty.
False, you may only move down
Average of the squared-differences in each individual measure value (x) and the groups’ mean.
Variance
Square root of variance value
Standard deviation
In a normal distribution curve the first standard deviation represents how much of the data?
68%
In a normal distribution curve the second standard deviation represents how much of the data?
95%
In a normal distribution curve the third standard deviation represents how much of the data?
99.7%
Statistical test useful for normally-distributed data.
Parametric (i.e. associated with interval/ratio data)
Data distribution, asymmetrical, mean > median, tail points right
Positively skewed
Data distribution, asymmetrical, mean < median, tail points left
Negatively skewed
Skewness value of 0.
Symmetrical data distribution (i.e. not skewed)
Positive skewness value.
Positively skewed
Negative skewness value.
Negatively skewed
Numerical value that evaluates how aggregated a data set may be (i.e. how clusted the data is)
Kurtosis (i.e. + = more cluster)
Four key questions for selecting the correct statistical test.
- What data level is being recorded? (i.e. nominal, ordinal, interval/ratio)
- What type of comparison/assessment is desired?
- How many groups are being compared?
- Is the data independent or related/paired?
Buzz words for correlation statisical test.
Correlation, association, relationship
Buzz words for regression statistical test.
Prediction, association, relationship.
Statisical test that derives a quantitative measure of strength and direction between the relationship of two variables.
Correlation
Nominal correlation test.
Contingency Coefficent
Ordinal correlation test.
Spearman Correlation
Interval/ratio correlation test.
Pearson correlation
A correlaiton that controls for confounding variables (i.e. only for interval data)
Partial correlation
A correlation test, shows relationship of agreement between/consistency of “decisions”.
Kappa statistic
+1 Kappa value.
Observers all classify everyone the same way.
0 Kappa value.
There is no relationship.
-1 Kappa value.
Observers classify everyone exactly the opposite of each other.
Statistical test, provides a measure of the relationship between variables by allowing the predictions about the dependent/outcome variable knowing the value/catergoy of independent variables.
Regression
Nominal regression test.
Logisitic regression
Ordinal regression test.
Multinomial logisitic regression
Interval/ratio regression test.
Linear regression
Statistical test, compare the proportion of events over time/time-to events between groups (ongoing progression)
Survival test
Common representation of survival tests.
Kaplan-Meir curve
Survival test for nominal data.
Log-Rank test
Survival test for ordinal data.
Cox-Proportional Hazards test
Survival test for interval data.
Kaplan-Meier test.
Buzz saying for survival test.
“Over time”
Statistical test for independent nominal data comparing two groups (i.e. for a normal to large set of data)
Chi-Square
Statistical test for independent nominal data comparing two groups (i.e. for smaller set of data).
Fisher’s Exact Test
Statistical test for independent nominal data comparing 3+ groups (i.e. from a normal-large data set).
Chi-square
Statistical test for independent nominal data comparing 3+ groups (i.e. from a smaller data set).
Fisher’s Exact
Post-hoc test, adjusts the p value for # of comparisons being made.
Bonferroni test of Inequality
Statistical test for paired/related nominal data comparing 2 groups.
McNemar Test
Statistical test for paried/related nominal data comparing 3+ groups.
Cochran
Statistical test for independent ordinal data comparing 2 groups.
Mann-Whitney test
Statistical test for independent ordinal data comparing 3+ groups.
Kruskal-Wallis test
What value/measurement is being compared in the ordinal group comparison tests?
Median values between groups
Statistical test for paired/related ordinal data comparing 2 groups.
Wilcoxon Signed Rank test
Statistical test for pair/related ordinal data comparing 3+ groups.
Friedman test
Post-hoc test, 3+ groups, ordinal data, compares all pairwise comparisons possible, and all groups must be equal in size.
Student-Newman-Keul test
Post-hoc test, 3+ groups, ordinal data, compares all pairwise comparisons againsts a singe control, and all groups must be equal in size.
Dunnett test
Post-hoc test, 3+ groups, ordinal data, compares all pairwise comparisons possible, and useful when all groups are not of equal size.
Dunn test
Statistical test for independent interval data comparing 2 groups.
Student t-test
Statistical test for independent interval data comparing 3+ group.
Analysis of Variance (ANOVA)
What value/measurement is being compared in interval data group comparison tests?
Means of all groups
Statisical test for paired/related interval data comparing 2 groups.
Paired t-test
Statistical test for paired/related interval data comparing 3+ groups.
ANOVA
Post-hoc test, 3+ groups, interval data, compares all pairwise comparisons possible, and all groups must be equal in size.
Student-Newman-Keul test
Post-hoc test, 3+ groups, interval data, compares pairwise comparisons against a single control, and all groups must be equal in size.
Dunnett test
Post-hoc test, 3+ groups, interval data, compares all pairwise comparisons possible, useful when all groups are not of equal size.
Dunn test
Post-hoc test, 3+ groups, interval data, compares all pairwise comparisons possbile, all groups must be equal in size, and slightly more conservative than the Student-Newman-Keul test.
Tukey test
Post-hoc test, 3+ groups, interval data, compares all pairwise comparisons possible, all groups must be equal in size, and less affected by violations in normality and homogeneity of variances (i.e. than Tukey or Student-Newman-Keul test).
Scheffe test (i.e. most conservative)
Statistical test, interval data, determines if variances are equal between groups.
Levene’s test
How can interval/ratio data be delt with when it is not normally distributed?
Transform to standardized value (i.e. z-score or log transformation) or drop down to ordinal data and use odrinal test.
Guide for determining whether we accept/reject null hypothesis.
Statistical tests
Not accepting the Null hypothesis when it is TRUE.
Type I error
Accepting the Null hypothesis when it is FALSE.
Type II error
Probability that differences in group is due to chance. Probabily of making a type I error had you claimed there was a difference.
P value
A studies ability to find differences between groups.
Power
True/False: More people in a data set increases power of a study.
True
Things to consider when determining sample size.
How small is the difference between groups being determined, expected variation of measurement, and Type I and Type II Error Rate or Confidence Interval.
