It’s been a while since a little burst of statistical fun hit the blogosphere but summer is sort of here, and you may be faced with a choice of tracks in a forest and unsure which one to take …

Path analysis will not aid you.

Path analysis (aka ‘structural equation modelling’ … sort of …) is a type of regression analysis, where you try to predict one thing that you don’t know (for example, final hight of a child) using stuff you do know (like the height of the parents, current age, height & pubertal status). 

‘Normal’ regression analysis says “how are these things weighted?”

‘Path’ style analysis sets up a hypothesis (final height depends on current age & height, but the power of these differ depending on puberty phase, and they are initially determined by parental height too) and then sees how strong those connections are.

What usually happens is you show that they aren’t very strong. (‘Cause the *real* path involves knowing the bone age, current IGF1 levels, nutritional status and duration of sleep nightly* and the path model works by assuming you know All The Things.) But they might help you understand suggested causality a little bit more.

• Bob Phillips

* everyone knows that you only grow in your sleep and that is why children sleep so much ’cause they need to grow