BODY SIZE AND MASS OF NGORONGORO CRATER LIONS

[WaveRiders]

tigerluver
 
My database is different and significantly wider, but probably overlapping in part with that one. For the regressions I have used to get those estimates I have not included my hunting records dataset.
 
Different database, perhaps different statistical methodology and therefore different results. Anyone should be confident in what he does. As a scientist and a professional I very much do, but I always know the limitations of what I am doing, accept the fact that I may do a mistake and always look to improve methodologies and results from my studies.
 
Chest girth is one morphological parameter that under certain circumstances can be well correlated with body mass. I remind you that in statistics correlation and accuracy (high prediction ability and low bias) is not the same concept. You can have regression with a very high correlation and a poor accuracy in the prediction. It would be too easy if chest girth predicted body mass within 1-2% for any individual. With one degree of freedom it is not possible to cover the individuality of an animal morphology where different body length, as you remind, and other proportions affect results. But this does not implicate that you have to use many variables. Among the many carnivores I have studied in depth only morphometry of polar bears typically adopts two parameters to predict body mass (chest girth and total length). Yet results are often debatable because of several factors like the seasonal variable adipose tissue deposit, variable morphological features of different population, age, sex and so on.
 
Among felids I can tell you that the main issue to get a well correlated regression with strong prediction ability for body mass from chest girth is get a database with weights adjusted for stomach contents. That makes a significant difference. Quite variable stomach contents is the primary reason why the guy who built the regression for lions and Amur tigers (and made a promising and good job) and that then did the same for Bengal tigers could not explain why the correlation for the latter from Cooch Behar Bengal tiger data was poorer. The lack of a good correlation and data spread across the regression is primarily, but not entirely, due to the variable stomach contents among the individuals of the dataset when using not-adjusted weights.
 
 
                         WaveRiders
 

 

  Reply