open-discussion
35 Subscribers 
open-discussion > Group test in Mancova
Showing 1-2 of 2 posts
Aug 18, 2011 12:08 PM | Onur Ugurlu
Group test in Mancova
Hi everyone,
When ı was doing the group test in Mancova, it didnt made any difference if ı selected "diseased 1, control -1" or "control 1, diseased -1" , the outputs files remain the same. So in the end the distances maps are the same for both of the two analysis.
How can I see in the distance map, if the diseased's parts or control parts have gotten any bigger or smaller ?
Any help will be grateful.
Onur..
When ı was doing the group test in Mancova, it didnt made any difference if ı selected "diseased 1, control -1" or "control 1, diseased -1" , the outputs files remain the same. So in the end the distances maps are the same for both of the two analysis.
How can I see in the distance map, if the diseased's parts or control parts have gotten any bigger or smaller ?
Any help will be grateful.
Onur..
Aug 25, 2011 03:08 PM | Beatriz Paniagua
RE: Group test in Mancova
Hi Onur,
For the way shapeAnalysisMANCOVA is coded, it does not matter what group numbers you select. You could have selected 100 and 9 as your group IDs and the program would have ended up standarizing those IDs. The important part here is that you are consistent giving group IDs, not really the number you assign. So the program is working as it should when you are switching your group ID labels.
About the distance map, if you used signed distances you could detect that there are negative and positive areas. These negative and positive values need to be put in context, i.e. even if you have negative values in one area that does not necessarily mean that you have shrinkage (it might be a displacement if you have positive values in the contrary border of your shape). If you are looking at absolute shape distance maps, you will see only shape diferences, regardless of direction. You will be able to dected which is which by seeing the value range of your data when loaded in visu (if you have negative values you are looking indeed at signed distance maps). Also, keep in mind that when is difficult to understand the distances computed by MANCOVA you only have to load the two mean meshes for your groups in Slicer or other program with multiple vtk visualization to reinforce your evaluation of the results.
Does it make sense?
Regards,
BEATRIZ
For the way shapeAnalysisMANCOVA is coded, it does not matter what group numbers you select. You could have selected 100 and 9 as your group IDs and the program would have ended up standarizing those IDs. The important part here is that you are consistent giving group IDs, not really the number you assign. So the program is working as it should when you are switching your group ID labels.
About the distance map, if you used signed distances you could detect that there are negative and positive areas. These negative and positive values need to be put in context, i.e. even if you have negative values in one area that does not necessarily mean that you have shrinkage (it might be a displacement if you have positive values in the contrary border of your shape). If you are looking at absolute shape distance maps, you will see only shape diferences, regardless of direction. You will be able to dected which is which by seeing the value range of your data when loaded in visu (if you have negative values you are looking indeed at signed distance maps). Also, keep in mind that when is difficult to understand the distances computed by MANCOVA you only have to load the two mean meshes for your groups in Slicer or other program with multiple vtk visualization to reinforce your evaluation of the results.
Does it make sense?
Regards,
BEATRIZ