Cross - sectional study design
- observational
- descriptive
- collects data from a population at one specific point in time
- groups determined by existing differences
Pearson’s correlation coefficient
measures linear relationship between 2 variables - p = 0 suggests no linear relationship
Pearson’s product-moment coefficient
examines whether continuous outcome variables are associated with a set of independent variables
Regression modelling
measure strength and direction of an association between variables
- continuous outcome - linear or non-linear regression models
- categorical outcome - logistic regression
Linear regression data considerations
Outcome variable (DV) must be continuous, independent variable (IV) can be categorical or continuous
Assumptions of linear regression
- relationship between DV and IV is linear
- observations are independent and randomly selected
- homogeneity of variances - constant variance
- residuals are independent and normally distributed
- absence of outliers and multicollinearity
Descriptives of outcomes
If normally distributed, skewness and kurtosis = 0
Mean, median and mode should be equal
Tests of Normality - MVPA
eg. Kolmogorov - Smirnov and Shapiro - Wilk
- non-significant test (P>0.05) suggests that the distribution is not significantly different from a normal distribution
- a significant test (P<0.05) suggests that the distribution is significantly different from a normal distribution
Multi-collinearity
refers to IV’s that are correlated with other IV’s
Variance Inflation Factor (VIF)
measure of how much variance of estimated regression coefficient is ‘inflated’ by the existence of correlation among IVs in the model
VIF 1 = no correlation
VIF >4 = more investigation
VIF >10 = serious multicollinearity
Constant Variance
Plot of fitted values against residuals - if scattered randomly around 0 - supports homoscedasticity
Can also use statistical analysis - if P>0.05 for homoscedasticity test - supports constant variance assumption
Breusch-Pagan test and Koenker test
- tests of heteroscedasticity
- if insignificant - can assume constant variances
- if significant - cannot assume constant variances
R^2
% of variability that is explained by fitted model
