Now a day, it is commonly perceived that if we apply complex mathematical or statistical models in our research then there are bright chances that research will be published. Whatever is your idea, you need to apply some models. How important is this concept?
Modelling is often necessary to summarize research results.Yes, the use of modelling enhances the publication of the results. The model should not necessarily be complex. The simpler a model is, the better it is understood and interpreted. This is the principle of parametric parsimony.
However the model should not also be an oversimplification.This means that some vital parameters should not be left out.
There are model selection criteria like the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Schwarz Information Criterion (SIC), etc. constructed such that each has an over-parametrization penalty and also an over-simplification penalty. There should be a balance.
Modelling is often necessary to summarize research results.Yes, the use of modelling enhances the publication of the results. The model should not necessarily be complex. The simpler a model is, the better it is understood and interpreted. This is the principle of parametric parsimony.
However the model should not also be an oversimplification.This means that some vital parameters should not be left out.
There are model selection criteria like the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Schwarz Information Criterion (SIC), etc. constructed such that each has an over-parametrization penalty and also an over-simplification penalty. There should be a balance.
Hi Anees,
Each fall I do power analysis to about 6-10 research design at our department and I must say that I don't agree with analysis complexity is a must for publishing research. Most of the studies conducted in my faculty are either trials or observational studies, and for power it is clear that model compexity is related to study design, that is the setting where the study is conducted.
On one hand we have randomized controlled trials (RCT) where it is possible to conduct analyses with very simple methods (t-tests, simple linear models), because they are designed in a rigorous manner and the design itself is used to remove bias from analysis. For this reason it is easy to also compute the required sample size for these studies. On the other hand, observational studies (OS) often lead to complex analysis situations, because of the need to consider effects that would otherwise cause systematic bias in making conclusions about the phenomenon under study. Accounting for such factors in a power analysis can be much more difficult.
In any case the researcher has to work hard on conducting the research, but while in the RCT study the heaviest work is carried out in the planning stage, the observational study must work harder on the analysis.
Regards,
Timo
The complex or simple models must be used correctly and the results must be reliable.
I dont think that complex models enhance the publication chance.
I have witnessed a situation when the use of too complicated mathematical (statistical) methods for modeling phenomena in other sciences often causes difficulties in the publication of results. For those who deal with a given phenomenon, mathematical methods make it impossible to understand the text and result in the rejection of work. On the other hand, mathematical journals are very often not interested in such results, because there is no significant mathematical content in them, and non mathematical problems are of no interest to the editors of mathematical journals.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling