One of my current interests is to investigate alternatives to the traditional null hypothesis significance testing approach for drawing inferences from research data. Many scientists are unaware of the myriad of problems with the use of null hypothesis testing in observational studies; therefore, I have compiled an extensive list of citations composed of papers outlining these deficiencies. I also have compiled a list of papers whose authors endorse, at least to a limited extent, the use of null hypothesis significance tests. However, authors are nearly unanimous in agreement that this approach is overused and misused in the scientific literature. I recently co-authored a paper with David Anderson and Ken Burnham in which we quantified this misuse/overuse in Ecology and Journal of Wildlife Management (pdf), as well as offered a more statistically rigorous alternative. Doug Johnson's 1999 invited article on the same topic, 'The Insignificance of Statistical Significance Testing,' was awarded the Outstanding Publication Award by The Wildlife Society. Those interested in exploring this topic further also should check out Bruce Thompson's webpage.
I am a strong proponent of the information-theoretic approach (AIC) to modeling and model selection as described by Ken Burnham and David Anderson in their excellent book, 'Model Selection and Inference: A Practical Information-Theoretic Approach' (1998, Springer-Verlag). I also strongly recommend Richard Royall's book, 'Statistical Evidence: A Likelihood Paradigm' (1997; CRC Press/Chapman and Hall) in which he discusses use of likelihood ratios as weights of evidence as a basis for statistical inference. Both approaches are based on R. A. Fisher's likelihood concept.