13
$\begingroup$

My interest in statistics is less on the math side but more on the use and misuse of objectively-valid statistics for political and economic purposes. Are non-math questions that involve issues of how to define data definitions, identify the appropriate number of demographic categories to use, or identity appropriate and meaningful boundaries among them in-scope at Cross Validated? For example, such questions might include, "Is it correct to include people who left high school early to complete a university degree in the definition of 'High School Dropout' for my research on literacy attainment?" or "If a person was previously incarcerated in a totalitarian country for purely political offenses such as criticizing the government and has now been convicted of their first crime in a 'Free' country, should I classify them in my statistical model of crime risk factors as first time offenders or recidivists?"

My guess is that neither of these questions would be in scope here and would be more suited for sites on Education and Criminology, respectively, or even a graduate classroom.

What is the consensus? Are these kind of questions welcome here or is the scope more or less limited to the math aspects of statistics?

Other potential questions along these lines:

  • I am trying to determine whether there is any correlation between the age at which a child completes toilet training and their likelihood to be injured by an electrical shock as an adult. If someone experiences a shock that is too weak to cause any physical damage but that causes the person to develop Post-Traumatic Stress Disorder, do I count that as an injury?
  • I am developing a model to find trends between the schools that a person has attended and the likelihood that they will be convicted of a crime involving misuse of computer technology. Does using an XBox as a doorstop during a residential burglary count as misuse of computer technology?
  • One of my research subjects says that he is a Mexican Chicano but adamantly denies being "Hispanic". Do I need to create a new "Non-Hispanic Chicano or Mexican" category for my model plotting ethnic background versus educational attainment or do Best Practices in demographic modeling allow me to define all Mexicans as ipso facto Hispanic?
| |
$\endgroup$
  • 4
    $\begingroup$ You have an impressively wide-ranging research program. $\endgroup$ – gung - Reinstate Monica Mar 9 '17 at 1:15
  • 1
    $\begingroup$ @gung they are just make-up examples to illustrate the kinds of questions I am talking about. $\endgroup$ – Robert Columbia Mar 9 '17 at 1:42
11
$\begingroup$

You can easily see that many of our questions are not mathematical and include questions about things like how to define variables (how many categories should I use? is a reasonably commonly asked question). Some of these questions - if appropriately framed - will probably be within the scope of our topic coverage, but for a number of those questions the chances of an answer that will be anything but vague may not be high -- in particular questions similar to "how many categories should I use" may be answered with anything from "don't split into categories" to "we can't answer that on the basis of this information" (which means some may eventually be closed as too broad or unclear without a lot more details), depending on whether it makes more sense to split at all.

The solution to that second kind of issue (how to categorize things given that you must do so) -- once you've identified the details enough to make it answerable at all -- will then probably already be about as obvious to you as to anyone else.

The more information you can give about your aims, specific questions of interest, preferences, intended audience, specific subject-area knowledge about the variables -- and so on (though preferably expressed succinctly) the more likely someone may be able to say something helpful. We generally can't just say "oh, use four categories". However, for many things, few categories will often be worse than many, unless other constraints force fewer on you.

So broadly, yes you can ask such questions, but answers that are simultaneously quick and useful may not be typical

As a specific example ""Is it correct to include people who left high school early to complete a university degree in the definition of 'High School Dropout'", "correctness" of your choice of what's included in a category is not a statistical issue, so that would need to be reframed to be on topic. (We can often answer questions about potential consequences of mixing two apparently quite distinct groups on your inferences, for example.)

Outside of such directly statistical considerations, whether it makes sense to include it or to do any of several other things really depends on what questions you want to answer, the way in which you want to answer them, the tolerance of your intended audience for doing so, and so on. This is not information we can give you (we don't tell you what you want to study and we're not typically experts in your application area), but instead you would need to supply any such information.

| |
$\endgroup$
11
$\begingroup$

In general, I think such questions are on topic. But my concern is that you may not get high quality answers at this site.

For example, political scientists may have totally different categorical definitions than economists, and statisticians can be completely ignorant of the fact that there's any difference between the fields. As such, I think that while these questions about can be on topic here, I think you'll have better quality answers about economic variables from an economic focussed forum.

To illustrate, I think this would be the right site for something like

'One of my research subjects says that he is a Mexican Chicano but adamantly denies being "Hispanic". What are the statistical ramifications for creating a new subpopulation with extremely few subjects?'. The question in bold is a definitive question about statistical implications.

The original question posed, "Do I need to create a new "Non-Hispanic Chicano or Mexican" category for my model plotting ethnic background versus educational attainment or do Best Practices in demographic modeling allow me to define all Mexicans as ipso facto Hispanic?", is likely to get a much better answer on a forum that focusses on demographic modeling specifically, especially in regards to what demographic best practices are.

| |
$\endgroup$

You must log in to answer this question.

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