6

Within the framework of an experiment I followed to growth rate of bird nestlings. I measured them every day for weight and tarsus. I have a number of continuous and categorical explanatory variables, as well as brood ID that should be used as a random factor. I am asking for advice, what would be the best way to analyse this data? In the case of the tarsus a sigmoid curve would describe the growth well, but in the case of body mass there is often decrease and no growth but reduction (see the graph attached).

I'd prefer using R, but any advice will be appreciated and considered.

enter image description here

amoeba
  • 93,463
  • 28
  • 275
  • 317
user3162369
  • 126
  • 6
  • 3
    You are looking for a nonlinear mixed effects model. You need to account for the fact that same chick was measured in time which is why the values are correlated. You can start [here](http://lme4.r-forge.r-project.org/slides/2009-07-01-Lausanne/8NLMMD.pdf). To get a list of models, check out `?lme4::selfStart`. – Roman Luštrik Jan 17 '17 at 17:11
  • Data cannot be simply "analyzed" because there are too many questions one could ask. Focus on a single question. Finding expected weight as a function of age is one question that everyone seems to think you are trying to ask. – AdamO Sep 17 '19 at 19:19
  • But I have more than that. For instance, the panel data are imbalanced. Do the curves stop when a bird dies or fledges? Did you measure those events? The "downward" trend could be predicted by season, death of parent, exogenous factors that are atypical "development" scenarios. – AdamO Sep 17 '19 at 19:21

1 Answers1

1

It would appear to me that you have daily time series data where you are measuring two characteristics. I would initially suggest using augmented ARIMA methods incorporating both memory and possible dummy indicators reflecting pulses. level/step shifts and/or time trends. Secondly it is possible to additionally relate these two series using a multivariate extension of ARIMA called Transfer Functions while investigating the impact of some of your user-suggested explanatory variables. If you wish to post data on one of your subjects showing both characteristics , I might be able to hep further. You can search SE for DAILY DATA or ARIMA and learn more about time series analysis. Model forecasts can always be converted to expected growth rates.

IrishStat
  • 27,906
  • 5
  • 29
  • 55
  • It might be helpful if the down-voter suggested a viable alternative or a pointed critique so that we could all become "smarter" . – IrishStat Jan 18 '17 at 14:05