13. Analysis of associations Part III: Non-Linear Regression

Question 1

what approaches are there to non-linear regression

Answer 1

• fit piecewise linear model
• polynomial model
• piecewise polynomial model
• transformation

Question 2

what is a piecewise linear model?

Answer 2

having multiple linear models to describe data

data is fitted between fixed points (knots)

Question 3

how do we compare quadratic models to linear models

Answer 3

anova

we need to check it it fits without assumptions too
- residuals are used
- done by comparing observations and predictions
- the plot of residuals should be random

Question 4

what non-normal approaches are there to non-linear regression

Answer 4

Count data (discrete data)
- Poisson distribution

Proportional data (fixed data range between 0 and 1)
- binomial distribution
- log-ods transformation to fit data , then back transform.

Survival data (fixed range between 0 & 1, measured over time)
- exponential distribution