Statistical tests (intro to biostats lecture) Flashcards
Student t-test
Interval
2 groups
Independent
Compares the means of all groups (along with intra and inter-group variations) against a dependent variable
(Pearson’s) Chi square (include its 2 assumptions)
Nominal
2 Groups
Independent
Compares group proportions and if they are different from that expected by chance
(usual chi-square data distribution and no cell count less than 5)
Fisher’s exact test
Nominal
2 groups OR more than 2 groups
Independent
Groups have an EXPECTED cell count of <5
(Bonferroni to adjust the p value for the # of groups being compared IF there are more than 2 groups being compared)
Chi square test of independence
Nominal
More than 2 Groups
Independent
(Bonferroni to adjust the p value for the # of groups being compared)
(usual chi-square data distribution and no cell count less than 5)
Bonferroni test of inequality (include the 3 tests run prior to the use of the bonferroni test)
Nominal
More than 2 groups
Independent OR related data
adjusts the p value for the # of groups being compared
(Chi squared of independence, Fisher’s exact, and Cochran are all tests that require the bonferroni as a “post hoc” test if they yield statistically significant findings [p value < 0.05])
McNemar test
Nominal
2 groups
Related
Cochran Q test
Nominal
More than 2 groups
Related data
Same principle/assumptions as chi squared, but it factors in related/paired data
(Bonferroni to adjust the p value for the # of groups being compared)
Mann-Whitney Rank Sum
Ordinal
2 groups
Independent data
Compares the median values between groups
(also used for interval data that does not meet parametric requirements)
Wilcoxson Signed Ranke
Ordinal
2 groups
Related data
Compares the median values between groups
Kruskal-Wallis Test (include what it needs if its value is statistically significant)
Ordinal
more than 2 groups
Independent data
Compares the median values between groups
(Student-Newman-Keul, Dunnett, or Dunn post hoc tests must be run to determine where there differences are when there are more than 2 groups)
(also used for interval data that does not meet parametric requirements)
Friedman Test (include what it needs if its value is statistically significant)
Ordinal
More than 2 groups
Related data
(Student-Newman-Keul, Dunnett, or Dunn post hoc tests must be run to determine where there differences are when there are more than 2 groups)
Levene test of equal variances
determines how “normally distributed” and the level of “equal variances” that are occurring in the data set
“it asks if the groups are equal”
paired t test
Interval
2 groups
Related data
Compares the mean values between groups that are related
ANOVA (include what it may need and why)
Interval
More than 2 groups
Independent data
Compares the means of all groups (along with intra and inter-group variations) against a dependent variable
(if this group comparison yields a statistically significant value, then a post hoc test must be conducted)(Bonferroni, Tukey, Scheffe, Dunn, Dunnet, Student-Newman-Keul)
ANCOVA (include what it may need and why)
Interval
More than 2 groups
Independent data
Confounders present
Compares the means of all groups (along with intra and inter-group variations) against a dependent variable while also controlling for the co-variance of confounders
(if this group comparison yields a statistically significant value, then a post hoc test must be conducted)(Bonferroni, Tukey, Scheffe, Dunn, Dunnet, Student-Newman-Keul)
Repeated measures ANOVA
Interval
More than 2 groups
Related data
Compares the means of all groups (along with intra and inter-group variations) of RELATED data against a dependent variable
(if this group comparison yields a statistically significant value, then a post hoc test must be conducted)(Bonferroni, Tukey, Scheffe, Dunn, Dunnet, Student-Newman-Keul)
Repeated measures ANCOVA
Interval
More than 2 groups
Related data
Confounders present
Compares the means of all groups (along with intra and inter-group variations) against a dependent variable while also controlling for the co-variance of confounders
(if this group comparison yields a statistically significant value, then a post hoc test must be conducted)(Bonferroni, Tukey, Scheffe, Dunn, Dunnet, Student-Newman-Keul)
Cox-Proportional Hazards test
Ordinal
Proportion with event (survival)
Compares the number of ordinal data events over time
Logistic regression
Nominal
Prediction or Association
(Predicts the outcome [dependent variable])
(you are able to calculate an OR for a measure of association)
Linear Regression
Interval
Prediction or Association
(Predicts the outcome [dependent variable])
(you are able to calculate an OR for a measure of association)
Multinominal Logistic Regression
Ordinal
Prediction or Association
(Predicts the outcome [dependent variable])
(you are able to calculate an OR for a measure of association)
Kaplan-Meier test
Interval
Proportion with event (survival)
Time itself is being evaluated
What can be represented by a kaplan-meier curve?
all survival tests
Log-Rank, Cox-Proportional hazards, and kaplan-meier test
Contingency Coefficient
Nominal
Correlation is desired
(partial correlation if you need to control for confounding)
Pearson Correlation
Interval
Correlation is desired
(partial correlation if you need to control for confounding)
A p value >0.05 with this test just means that there is no “linear” correlation, and there may still be non-linear correlations present
Log Rank Test
Nominal data
Proportion of events (survival)
Compare that proportion of nominal data value over time
Spearman correlation
ordinal
Correlation is desired
(partial correlation if you need to control for confounding)
Student-Newman-Keul test
Student-Newman-Keul test: compares all pairwise comparisons possible and all groups MUST be equal in size
Dunnett test
Dunnett test: Compares all pairwise comparisons against a SINGLE CONTROL and all groups must be equal in size
Tukey test
Tukey test: compares all pairwise comparisons possible and all groups MUST be equal in size
Slightly more conservative than the Student-Newman-Keul
Scheffe test is less affected by
Dunn test
Dunn test: compares all pairwise comparisons possible and is useful when all groups are NOT of equal size
Scheffe test
Scheffe test: compares all pairwise comparisons possible and all groups MUST be equal in size
Scheffe test is less affected by violations in normality and homogeneity of variances - most conservative
