Module 6 Flashcards
correlation anaylsis
uses direction (positive or negative) and magnitude (-1, 1 perfect correlation, 0 no association) of a relationship
pearson correlation (r) assumptions
assumptions:
- scale
- paired data
- normality
- no outliers
- linearity
- homoscedasticity (equal variance)
checking for linearity and homoscedastiticy
use scatter plot, not equal variance is plots fan in or out
pearson correlation non parametric
spearmans correlation (rs)
intra class correlation coefficient (ICC)
inter rater reliability - variation between two or more raters measuring the same event
test retest reliability - variation in two measurements from same instrument, subject, rater and conditions
intra rater reliability - variation within one rater across two pr more trials
ICC level of agreement
1 excellent, less than 0.4 poor
factors impacting correlation
- restricting data range, can remove non linear portions
- heterogenous (natural variability among data), sample can mask patterns
- outliers
regression (r^2) + assumptions
investigates how one variable influences another
- response variable - predicting (dependent)
- explanatory - using to predict (independent)
assumptions:
- scale
- normality
- no outliers
- linearity
- homoscedasticity
epsilon
best fit line error term
- represents residuals (distance of observed Y from line)
least square method
minimizes sum of squared residuauls
- sum must = 0
- line of best fit passes through mean of X and Y
multiple linear regression
one response variable, more than one explanatory variable
logistic regression
response variable is dichotomous
statistical significance
quantifies probability of an event occurring due to random chance
clinical/biological significance
event or difference is meaningful for a clinical/biological reason
causes of replication crisis
- publication bias
- study design and power (few observations)
- questionable practices (selective reporting/sampling bias)
- HARKing (hypothesis after results)