Tuesday, November 17, 2009

Models and muddles the divergence of science from theory.

It is inevitably necessary to think of all as contained within one
nature; one nature must hold and encompass all; . . . But within the
unity There, the several entities have each its own distinct existence.

Nalimov Faces of science p136

Models are abstracts of the real world (and are bounded by our understanding of natures laws).The models are formally experiments or measurements that unfold our understanding of real world process.If a model or experiment is published that violates natures laws and yet correlates with the experiment,we are left with a simple binary question (and answer).That either the model(experiment) is incorrect or the natural law is wrong.

One would think that this would be a significant barrier to publication in a kournal such as nature or science,but this seems not to be the case.

Ontogenetic growth: models and theory

Anastassia M. Makarieva, Victor G. Gorshkov and Bai-Lian Li

We re-analyze the assumptions underlying two recently proposed ontogenetic growth models [Nature 413 (2001) 628; Nature 417 (2002) 70] to find that the basic relations in which these models are grounded contradict the law of energy conservation. We demonstrate the failure of these models to predict and explain several important lines of empirical evidence, including (a) the organismal energy budget during embryonic development; (b) the human growth curve; (c) patterns of metabolic rate change during transition from embryonic to post-embryonic stages; and (d) differences between parameters of embryonic growth in different taxa. We show how a theoretical approach based on well-established ecological regularities explains the observations where the formal models fail. Within a broader context, we also discuss major principles of ontogenetic growth modeling studies in ecology, emphasizing the necessity of ecological theory to be based on assumptions that are testable and to be formulated in terms of variables and parameters that are measurable.

Evidently a problem but the continued repeating of the error in subsequent papers,suggests a significant failure of the peer review process.

eg Anastassia Makarieva

The validity of the fundamental laws of nature and of good theories based on them
has been tested on such a great amount of empirical data that it is a good theory that can tell you whether the empirical data are of good or bad quality rather than the data tell you something about the theory. For this reason, good theories can be used for making predictions, like the existence of many elementary particles was predicted in theoretical physics prior to their actual discovery. How justified is the use of models for making predictions?

During model development the priority is given to reaching a satisfactory agreement
between the data and the mathematical structure of the model. On the basis of theavailable sets of data points taken from the general ensemble of all empirical evidence the modelers determine linear and non-linear correlations between the chosen measurable variables, including their temporal changes. The resulting time dependence of model variables allows one to make a forecast for the future. Such a forecast, however, is nothing but a limited extrapolation of what has been observed in the past. With changing the empirical datasets the model structure and forecasts change. With inclusion of ever growing amounts of observations the models become more and more complex, while their agreement with the available observations naturally improve. Thus, an ideal model ultimately comes as an exact and convenient, i.e. mathematically formalized, representation of all the available data. However, to the degree the model is a model and not a theory, it lacks the predictive power. Because of the obvious fact that it cannot be expected that the calibrations made on the basis of the knowndata will remain valid in the domain of predicted (i.e. still unknown) data. This is a conceptual, fundamental problem with the modeling approach. The universal laws of nature predict things

Based on our own scientific expertise, we can illustrate the above points with specific examples of models that were judged to be most successful based on their agreementwith the data and claimed derivability from a "universal" theory, yet shown to confront the fundamental laws of nature. As one can see, the problem transcends across the natural science as a whole. The biological model of organismal growth (West et al., 2001) misinterpreted the energy conservation equation and replaced it with the one conflicting with the energy conservation law. Despite that, the model showed perfect agreement with the data. After the error was identified (Makarieva et al., 2004) it took the model’s authors four years to explicitly admit it (Moses et al., 2008) and re-formulate the model. The re-formulated model re-calibrated using the same data as the original (wrong) one showed equally good agreement with the data and got equally well published (Hou et al., 2008). Thus, irrespective of conflicting with the energy conservation law or not, the model agreed with the data, was widely cited and raised little concern in the reading audience

One could reflect on Alven Tofflers suggestion that the illiterate of the future will not be an inability to read and write.but to learn and think.


Post a Comment

Links to this post:

Create a Link

<< Home

Web Counters