(I have asked this question originally on Cross Validated; however, no good answer and someone suggested me to ask the question here).

We know that we can perform statistical test (e.g., t-test) for two sets of measurements (e.g., some distance measurement) stored in vectors A and B easily.

However, currently I want to perform a similar statistical test for two sets of points (for example, with x-y coordinates stored in A and B). May I ask how to do this in Matlab?

Here is a picture showing what I want to do: enter image description here

'A' stores the coordinates of the red stars, and 'B' stores the coordinates of the blue circles. These points will be some manual annotations from two human operators. The idea is trying to use statistical test to test whether the annotations from these operators are significantly different or not.

N.B. The points are not 1 to 1 mapped. For 1 point in set A, it can have several corresponding point (nearest ones) in set B.

  • @PolyGeo Someone suggested me to migrate my question to the GIS forum. I am not sure how to do that because I don't think I have any admin privilege. If possible, please close that question, and I would like to see if anyone here can help. Thanks very much. – Samo Jerom Feb 2 '16 at 11:54
  • You could have flagged your question on Cross Validated and asked a moderator there to migrate it here - we cannot pull it, they have to push - now that you have already posted here I think your only option is to delete it from Cross Validated but I am not sure that your question is on-topic here - let's see what the community thinks. – PolyGeo Feb 2 '16 at 11:59
  • @PolyGeo Thanks for your tolerance. I have already deleted the Cross Validated one. For the Mathematics post, I asked a slightly different question about combining metrics. If people think this question is not suitable for GIS, please close or delete this later. – Samo Jerom Feb 2 '16 at 12:05
  • How many corresponding B points can be mapped to one point A? There should be only 1 nearest or do you want to compare it to the nearest n points in B? – Kersten Feb 2 '16 at 13:16
  • @Kersten Thanks for your reply. What I mentioned is that there is no one-to-one mapping between A and B. Theoretically, there can be a situation that two points (or more) in A have exactly the same (minimum) distance to one point in B. But for my problem I think I need to find the nearest one, not nearest n points. – Samo Jerom Feb 2 '16 at 14:32

Maybe you can use PAST software to add your data as two-dimensional landmark coordinates, perform general Procrustes analysis and then a Principal Component Analysis ?

  • 2
    Could you elaborate a bit ? How can the asker perform a procruste analysis ? What does this analysis returns ? It is an indicator on similarity ? What would bring a PCA ? I added links to your answer to help understanding – gisnside Oct 9 '17 at 7:01

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.