Biostats Flashcards
Null Hypothesis (Ho)
-research perspective that states there will be no (true) diff b/w the groups being compared!!!
statistical perspectives
- superiority
- non-inferiority
- equivalency
mode, median, mean, outliers
- mode- most freq/common
- median- middle
- mean- avg
- outliers- impact the MEAN!!
inter-quartile range (IQR)
-middle 50%
Shapes of data distribution- symmetrical
(bell-shaped)
-normally distributed!!!
std deviations
- 1 std deviation above/below mean- 68% of pop
- 2- 95%
- 3- 99.7%
Positively Skewed
- tail points to the right
- mean > median
Negatively skewed
- tail points to left
- mean < median
Skewness- measures?
- measure of the asymmetry of distribution
- a perfectly-normal distribution is symmetric and has a skewness value of 0
3 primary levels for data
- nominal
- ordinal
- interval
2 key attributes of data measurement
- Magnitude
- fixed interval
Nominal
- dichotomous/binary (2 categories); non-ranked named categories
- NO magnitude, NO interval
Ordinal
- ranked categories
- YES magnitude, NO interval
Interval
- units!!
- YES magnitude, YES interval
required assumptions of interval data
- normally distributed
- equal variances
- LEVENE’S TEST- when want to know if variances or spread of groups are equal!!
Handling interval data not normally distributed
-cant be interval!
4 key questions to select the correct statistical test
- Data level- magnitude, interval
- type of comparison
- how many groups- 2 or 3 or more
- data independent or related (paired)
type of comparisons
- correlation
- regression (prediction)
- survival comparison (time)
- group comparison
Correlation (r)
- quantitative measure of strength and direction of a relationship b/w variables
- values range from -1.0 to +1.0
- +1.0- perfectly positive- 45 degree angle, positive slope
- 0.0- no association
- -1.0- perfectly negative- 45 degree angle, neg slope
Correlation tests
- nominal- contingency coefficient
- ordinal- spearman correlation
- interval- pearson correlation
Survival tests- represented by?
(TIME)
-kaplan-meier curve
Survival tests
- nominal- Log-rank test
- ordinal- cox-proportional hazards test
- interval- kaplan-meier
Regression tests
(prediction)
- nominal- Logistic regression
- ordinal- multinomial logistic regression
- interval- linear regression
Nominal data- 2 groups of indep data; 3 or more groups of indep data; 2 or more groups w expected cell count of <5
- (Pearson’s) Chi-square test- 2 groups of indep data
- Chi-square test of independence- 3 or more groups of indep data
- Fisher’s Exact test- 2 or more groups w expected cell count of <5
Nominal data- 3 or more groups of indep and related data- post-hoc testing
-Bonferroni test of inequality
Nominal data- 2 groups of paired/related data, 3 or more groups of paired/related data
- McNemar test- 2 groups
- Cochran- 3 or more groups- Bonferroni test of inequality
Ordinal data- 2 groups of indep data, 3 or more groups of indep data
- Mann-Whitney test- 2 groups
- Kruskal-Wallis test- 3 or more
Ordinal data- 2 groups of paired/related data, 3 or more groups of paired/related data
- Wilcoxon Signed Rank test- 2
- Friedman test- 3 or more
Ordinal data- 3 or more groups of indep or related data- post-hoc test
- Student-Newman-Keul
- Dunnett
- Dunn
Interval data- 2 groups, 3 or more groups of indep data, 3 or more groups w Cofounders
- student t-test- 2
- ANOVA (analysis of variance)(1 DV), MANOVA (mult analysis of variance)(2 or more DVs)- 3 or more
- ANCOVA, MANCOVA- w cofounders
Interval data- 2 groups, 3 or more groups of PAIRED data, 3 or more groups w Cofounders
- paired t-test- 2
- repeated measures ANOVA, repeated measures MANOVA- 3 or more
- ANCOVA, MANCOVA- w cofounders
Interval data- post-hoc tests for 3 or more
- student-newman-keul test
- dunnett test
- dunn test
- turkey or scheffe tests
- bonferroni connection
Kappa statistic
- correlation test showing relationship or agreement b/w evaluators (consistency of decisions)
- +1- observers perfectly classify everyone exactly the same way
- 0- no relationship at all b/w the observers’ classifications
- +1- observers classify everyone exactly the opp of each other
type 1 error
-rejecting Null hypothesis when it is actually TRUE, and you should have accepted it!!
(claim a diff when there isnt 1)
type 2 error
-not rejecting null hypothesis when it is actually FALSE, and you should have rejected it
(when there is a diff)
Power (1-B error)
- ability of a study design to detect a true difference if 1 truly exists
- sample size- inc power!!!
p value
- probability of making type 1 error~
- <5%- can reject the null hypothesis- statistically significant
- > 5%- not statistically significant
Confidence Interval (CI)
- 95% confident that the true diff or relationship b.w the groups is contained within the confidence interval range
- if CI crosses 1.0- NOT significant (p>0.05)