[R-sig-ME] Mixed-level regression vs. Box-Jenkins time series analysis

[Hein van Lieverloo]

Dear group,

 

Can anyone help me find information to compare applicability of:

*	Mixed-level regression analysis with glmmTMB, with repeated measures to deal with autocorrelation. The predictor of the effect of a change could be: 0 (before) and 1 (after).
*	Box-Jenkins time series analysis (designed to deal with autocorrelation) with multiple predictors, a.o. effect of change (comparing before and after).

 

Background of this question: Some biological cases I’m working on in the drinking water industry in The Netherlands:

*	for many years (in my spare time, which is limited): glmmTMB: Mixed-level zero-inflated Poisson-regression: invertebrate counts in drinking water distribution mains (level 1: distribution systems of 34 treatment plants, L2: 176 sampling hydrants, L3: 8 quarterly samples (time series), a multitude of variables. Seasonality is limited. I think Box-Jenkins is not the solution here.
*	with a colleague of mine, we try out the Box-Jenkins approach (I’m not familiar with it) with his program Time Series Analyst (self-developed in MatLAb). I think this can be done with glmmTMB as well, but what is best?  Simple methods (Mann-Whitney, multiple regression) are affected by single predictor issue and effects of zero counts respectively. And in any case, we want to know what method to use to take autocorrelation into account.

*	Surface water system: time series of predictors (bird species counts and temperature (seasonal) and a change in infrastructure) on response (bacterial counts).  Q: does the infrastructural change affect bacterial counts?
*	Drinking water distribution system (urban area): time series of nominal predictors (change in water quality entering whole system, one part of the system cleaned, one part not cleaned) on respons (bacterial counts). Q: what is the effect of the water quality change and what is the effect of cleaning?

 

FYI: all these issues are not a public health problem, mainly scientific research.

The Netherlands water companies boast having the best drinking water in the world, without a disinfectant residual (I have not found arguments to disagree yet)..

 

Kind regards,

 

Hein van Lieverloo

Viaeterna

 

 

J.H.M. van Lieverloo

E-mail: hein.van.lieverloo using viaeterna.nl <mailto:hein.van.lieverloo using viaeterna.nl>  

Mobile +31 6 24269886

www.viaeterna.nl <http://www.viaeterna.nl> 

 

 


	[[alternative HTML version deleted]]