We previously had a question which was deleted which was as follows:

Classification of RT?

Hope this is the right place for this question: I have an experiment of two conditions & I have the reaction time (RT) for those two conditions (180 subjects and 500 RTs) the mean RT is significantly differ (t-test) between the two conditions. Now, I would like to construct a classifier which will learn the differences between those two conditions and will be able to classify new instances. I guess some dimension reduction is needed ? which classifier is recommended in such case ? Does anyone aware of such kind of work ? Any information on this would be appreciated.

The question has a psychological variable (i.e., reaction time) in the question, but the focus is on statistical analysis and machine learning.

Thus, in broad terms what should be done with questions that overlap psychology and statistics?

share

2 Answers 2

Migrate:

  • If the psychology variable is of minor relevance to the question, and it's really just a question about statistics then I'd migrate to http://stats.stackexchange.com/ . A test for minor importance might be that answers would not be particularly influenced were the variables changed to variables from a completely different discipline.

Don't migrate

  • If the question relies on content knowledge in psychology or cognitive science, even though it has a strong quantitative component then I'd let the question stay here;
  • Stats.se might provide a good answer. However, there is general assumption that some questions fit on multiple sites and it's up to the person asking the question to decide in that case where they want to ask the question.
  • As a secondary point for borderline cases, if the question gets a good answer here, then I see less need to migrate.

On topic examples

These are some examples of topics that seem to me to be on topic on cogsci

  • Questions about processing reaction times
  • Questions about estimation of mathematical models that are specific to the cognitive sciences
  • Psychometrics, scale construction, scoring tests, norms, IRT, etc.

Specific examples

  • With regards to the question that motivated this meta post about RT classification, I think it's a borderline case. I think the question would get a better answer on stats.se and probably should have been migrated there.
share

The main problem in my opinion is the duplication of effort, lack of exposition to the right audience and potential for confusion created by the overlap.

For example, questions about psychometrics could in principle be on-topic here if there were no other StackExchange site for them but there are already many of them on CrossValidated and a few very knowledgeable users there.

CrossValidated is older, bigger and has more users that could be able to provide useful answers. Systematically migrating borderline questions would ensure that these resources are used to the fullest possible extent whereas scattered questions are less useful for everybody, less likely to be found (if you come across CrossValidated, will you think about checking if there is a “cognitive science” website?) and less likely to attract users interested in answering them.

I therefore believe that we should be a little more aggressive in migrating stat-related questions.

share

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .