What if you used an older paper and based your model on it, but authors published a newer paper with modifications you neglected?

Question

Is it ok to form a model based on several papers and omit something from newer papers? Of course, not that someone else proved what you're researching, just that authors said that something was important for one segment and you investigated another segment and neglected that. Later, the author published something and considered the thing you neglected. For the scope of research you couldn't include that and you based your argument on research context, scope and author's first publication.

What do you think? Is this a problem? Does it happen that authors omit some information because it doesn't help them? Please share your thoughts with me. Thanks.

Furthermore, is it ok to make changes in your argument after peer reviews even if reviewers didn't notice and didn't consider it a mistake.

Since I'm new in this research field, I have so many questions. I hope you understand.

Answer 1

Furthermore, is it ok to make changes in your argument after peer reviews even if reviewers didn't notice and didn't consider it a mistake.

No, absolutely not.

Unless you mean change your argument and then submit for peer review again.

In that case the answer would be ... still "No".

It is not okay. It is a waste of time: you are your first peer reviewer, do at least a good job that you and yourself are agreeing with your arguments, before sending it to other reviewers.

Answer 2

It depends on the purpose of your research. As G. Box said "all models are wrong, but some are useful"; wikipedia entry. For example, suppose you are modelling a biological phenomenon. Then, you'd use the simplest model that suits the desired complexity and precision (see also Occam's razor and bias-variance trade-off). If you aim to model every tiny effect, then you'd need a large, all-encompassing model; if you're focusing on the main effects, then a simpler model -- often a subset of the former -- would suffice.

So, the earlier model published may be a useful subset of the authors' later model and might suffice for your research purpose and empirical data. What is a no-no, however, is to cherry-pick data to fit your model and discard counter-evidence without a good explanation. Another possibility is that you are writing a methodology paper, to extend the model itself. If this gives an original and useful extension, then why not use only the earlier models, if they are enough? There could be details: for example, if the earlier models have shortcomings, you'd need to address these.

(Note: There appears a separate question in the penultimate paragraph of the original post; this is addressed in @EarlGrey's answer.)