2

I decided to make an archaeological map about certain tombs distributed in the area of Crete and make some statistical analyses using the spatial statistical tools of GIS.

enter image description here

Specifically, I'll be using the nearest neighbor analysis to find out whether they are clustered and a Kolmogorov-Smirnov test to examine their relation to their elevation. I was faced with a dilemma - for several tombs the coordinates were exactly the same. Now this could have happened because they were constructed very close to each other and the GPS did actually capture the same numbers or the archaeologist who did the survey thought he should just take one point for the whole cemetery.

How can I deal with this problem?

Would it be wrong if I performed the test using more than once the same coordinates?

The maximum is three tombs having the same coordinates.

  • Could you explain how the K-S test could be applied to "examine a relation" between locations and elevations? It is not normally considered for such applications, because it was developed to compare a univariate distribution of data to a reference distribution. It's hard to see how such a comparison could tell you much about location-elevation relationships. For clustering analysis, why does it matter that some locations coincide? What clustering method(s) are you using that would be sensitive to that? – whuber Aug 22 '13 at 21:22
  • The answer to your question depends on the full size of the analysis area upon which you will be running your analysis. If the analysis is done over a huge area and the size of the cemetery is but a tiny blip on the map, then the cluster analysis differences will be insignificant. If, however, the cemetery in question is 50% of your analysis area, then you are going to have a problem. – Conor Aug 22 '13 at 21:24
  • @whuber Thanks for your answer, I am trying to find out whether there was a certain pattern in the choice of the tombs' construction. So, yes I am testing the relation between the locations of the tombs and their elevation. They are not ordinary tombs they are very big structures and they might be also territorial markers... – Anna Xilakis Aug 22 '13 at 21:33
  • @Conor the area is not huge nevertheless it is the area where these tombs appear so it's not random. The results showed that they are clustered. Would it be wrong, or is it a totally relative matter? Thank you for your answer :) – Anna Xilakis Aug 22 '13 at 21:34
  • 1
    Regarding the post above, we cannot answer your clustering question without more information. Maybe posting a screenshot of the points with a scale bar would provide more insight. – Conor Aug 22 '13 at 21:55
2

I would just nudge the stacked points off of each other by a small margin. A meter or two shoudn't make a huge difference in your analysis when the relative variables are more likely at the scale of 10s of meters or more.

I can't offer any substantive advice as per the statistical concerns raised in the comments above. However, as an archaeologist and GIS professional I see no problem with simply adjusting the point locations to avoid having them stack right on top of each other. After all, something like a tomb is not terribly well represented as a point anyhow.

  • thank you for your response, since you are an archaeologist yourself how would you deal with the problem of the points representing the tombs? Would you do something like drawing a circle based on the diameter of each tomb maybe? – Anna Xilakis Aug 22 '13 at 22:59
  • First off, I'm an archaeologist in North America, I have no experience with European archaeology. That said, I would think that you could use the site reports and maps to digitize the area for each site, perhaps even fairly accurately. If you don't have the maps, then some manner of polygon that relates back to the size described in the report would likely suffice. – Kevin Aug 23 '13 at 13:34
  • 2
    My point is that archaeological sites always have area. Points never do. So, by using a point to represent an archaeological site you may be incorrectly limiting the number of other variables that your site is associated with. For example, the relevant variable may be a range of elevations, not a single elevation, and by assigning each site a specific elevation from a point you are obscuring that. Choosing to use points or polygons is a choice that should include consideration of the scale of your analysis, your research question, and what other variables you are considering. – Kevin Aug 23 '13 at 13:41
  • These are the instructions I had to follow, I do not really have a choice at the moment but your comments were very helpful for future projects :) – Anna Xilakis Aug 23 '13 at 15:23
  • 1
    Totally understandable! To solve your original problem, I'd stick with my first suggestion. Just nudge them off to the side a bit. If it were my project, I'd move each one 1-2 meters away from the original location, at least that way they aren't stacked directly on top of each other. – Kevin Aug 23 '13 at 15:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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