My question is both conceptual and technical, but more conceptual than technical. It's a little long and multi-pronged, but I figured the best answers would consider everything, so thanks for bearing with me. I list specific questions later.

Right now I'm research the concept of "academic mismatch" in undergraduate education, or when students attend colleges where their academic background is substantially different than the typical student (e.g., a "little fish in a big pond:" a student with low grades/scores compared to his peers, or a "big fish in a little pond:" a student with high grades/scores compared to her peers). While I'm on solid footing with outcomes analysis (do mismatched students do better/worse in school?), I want to explore the mechanisms of attending a school as a mismatch in the first place. There are a lot of reasons for this, some obvious and well explored (preferential admissions policies), others less obvious and less well explored (where I hope to contribute).

One question I want to explore, and many others (including my dissertation advisor) have found interesting, pertains to where students live and their likelihood of mismatching. The basic hypothesis is that when a student has fewer colleges to choose from within his "local college market," the more likely he is to attend a college that is less selective than he could because it's convenient. Likewise, for non-selective schools, students sometimes attend less academically appropriate colleges because they can get in, even if they're outmatched by their typical peers. There's some evidence out there in local contexts (e.g., Chicago), but very little in a national context, and almost none in rural contexts (which I/we find most fascinating).

This is the basic question I want to answer:

Does the concentration of "good-fit" colleges that are geographically near to a student associate with that students likelihood of attending a college where her academic background matches her peers?

So from this, I need to do the following:

  1. Create a measure of the concentration of good fit colleges. I have methods for deciding if a college is a good fit between student and college, but assessing concentration is more challenging.
  2. Test whether this measure is associated with students "mismatch status"
  3. Create maps demonstrating patterns found

Here are my questions:

  1. My intuition for concentration is to simply count the colleges that are a good fit within a specific distance (or a market defined by a shapefile) for each student and do appropriate analysis. What better ideas do you have?
  2. Are there geospatial statistical tests that are better than my plan (logistic regression analysis with outcome being match/not match, independent variable being the measure of college concentration?
  3. Software wise, should I bite the bullet and learn ArcGIS? I have experience as a web developer so my intuition is to run to web-based software, but I'm not sure where to start. My problem seems like it could be solved without ArcGIS, and if I can learn software I can use in web development, all the better.

As background, I have the following data:

  1. A student's home zip code and a number of variables on their academic background (high school gpa, SAT, etc.)
  2. A data file of all accredited colleges and their academic profiles (which I can assign geo-coordinates)

Everyone is in the USA.

I have time to learn, but am in proposal-writing stage, so insights in that regard are especially helpful.

As background, I have basically zero experience with GIS, but very good statistical analysis, computer programming skills, and data manipulation skills (for a social science researcher). Put another way, I'm comfortable with research design and Stata, R, Python, MySQL, but have never done anything with GIS (besides some web analytics using geolocation with IPs).

closed as too broad by PolyGeo Jan 10 '15 at 8:03

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 2
    Welcome to GIS SE! Your proposal sounds commendable but unfortunately I think it is too broad to fit within the scope of GIS SE. Consequently, I recommend that you try to focus your Question down to the most important question you want help with and then research/ask the other questions separately. Our protocols take some getting used to but I think this is a good read on what makes a good GIS SE Question meta.gis.stackexchange.com/questions/3349/… PS Yours is definitely NOT a terrible Question so don't be put off by the opening statement in that URL. – PolyGeo Jan 22 '14 at 0:59
  • 1
    Some excellent questions there - though you could do a lot without strictly resorting to a spatial analysis in GIS (once you have your measures of distance / concentration which could even be ordinals or classes). As suggested, click edit and pare this down to a single question, then ask others separately! – Simbamangu Jan 22 '14 at 4:57

I found your proposal is quite interesting! I have some personal comments here as I think it might provide some ideas for your research.

From your post, it looks like you intend to compare distribution patterns behind a set of geographically abstracted points, which can be broadly categorised into two distinct groups (i.e. the locations of colleges, and the home locations of students). In order to find out some "good-fit" stories among them, you attempt to find out their spatial and attribute relationships by using GIS and statistical analysis tools. With regard to point concentration analysis, this concept in GIS is often named as "hotspot detection" or "hotspot analysis".

There are many spatial methods, as I know, that have been explored in the past, which can be used to identify the "true" location of their concentration centres, e.g. kernel density estimation, ellipse based method, etc., rather than a specific geographical distance. It is not quite easy to explain all of them here but I can recommend a paper for you, and it can provide an overview of these methods and their applications.

McCullagh, M. J. (2006). Detecting Hotspots in Time and Space, University of Nottingham.

In addition, some spatial tools like spatial autocorrelation, K function analysis, I guess they should be useful for your study to determine whether the spatial relationship is random or relevant.

From software point of view, there are quite a lot of small, useful spatial programs that have been released beside ArcGIS, e.g. CrimeState (http://www.icpsr.umich.edu/CrimeStat/), (you can find others based on the paper above and relevant links). Many point analysis tools are available for you to use in CrimeState. However, I think a GIS platform is still required (e.g. ArcGIS, or QGIS to present data spatially on Desktop).If you have programming skills, some web GIS library can be used for you to display data online after analysis, e.g. OpenLayers, SharpMap, GeoServer, etc.

Not the answer you're looking for? Browse other questions tagged or ask your own question.