Basic statistical principles Flashcards
definition: statistics that assumes a normal distribution (bell curve) and needs a large # of subjects (n > 30)
parametric stats
definition: statistics that does not require a normal distribution or if the distribution is unknown (usually has n<30)
non-parametric stats
definition: the probability that the findings occurred due to chance
p values
definition: the range of scores in which the true score for the variable is estimated to lie within a specific probability
confidence interval (CI)
definition: the probability that a statistical test will detect a relationship between 2 or more variables OR a difference between 2 or more groups of scores
power
What % of power is common in most studies?
80%
definition:
- Rejecting the null hypothesis when a relationship does NOT exist
- Saying there is a difference when there is not
type I error (false positive)
definition: Accepting the null hypothesis when a relationship DOES exist
type 2 error (false negative)
definition: “there will not be a difference between the two groups”
null hypothesis
definition: “ there will be a difference between the two groups”
directional hypothesis
definition: A test used when you are trying to determine differences
- Assuming normal distribution with population size >30
independent t-test
definition: A test that compares the two populations with the subject being the control
EX: balance test/timed test (4-square test) before therapy vs. after therapy
paired T-test
(true/false) you can use a t-test when there is more than one group
FALSE
With ANOVA (parametric stat), you have to use a _____ test if there is statistical difference between the groups due to ANOVA not indicating which group had a difference
post hoc
definition: Similar to ANOVA but controls for or a co-variable, only 1 dependent variable
ANCOVA
definition: parametric stat that has more than one dependent variable/outcomes
EX: comparing balance and strength between low intensity vs. moderate intensity. vs high intensity
MANOVA
What are non-parametrics stats used to test differences?
mann-whitney U
wilconxon Rank Sum
Wilcoxon Signed Rank Test (similar to paired t-test)
kruskal-wallis H (similar to ANOVA- compares 3+ groups and required post hoc test)
Freidman’s ANOVA (ANOVA w/ repeated measures, requires post hoc)
definition: difference between the mean of the 2 groups examined / SD
effect size
What is considered as a good effect size?
0.8 +
What is considered as a bad effect size?
0.2
