Hi Onur,

In Mancova the rule of thumb is to see your variables divided in two categories: categorical, and linear. Categorical are variables that divide your sample in groups, i.e. female vs male, diseased vs control, etc etc. These are usually used in group studies. Linear are linear values assigned to each of your samples i.e., the level of a protein in blood for that sample, behavioral scores, pain scores, etc. These are usually included in interaction studies.

So with regard to your questions...

1) Yes you can use gender in your analysis.  So if you want to add gender to your analysis just add a new column giving a categorical value to all your sample (female = 0, male = 1). Then do a group study.

2) With regard to longitudinal changes in Mancova, we dont usually use the pure raw 3d information that we get with the SPHARM meshes. For that we compute displacement vectors and input those in as input data in Mancova (there will be a colum with the path to your *txt vector files as input). The same analysis that can be done to meshes can be done with displacement vectors. To compute displacement vectors you can use MeshMath (included in the ShapeAnalysisModule distribution). 

MeshMath MetaMeshT2 -subtract MetaMeshT1 will compute absolute distances and vector displacements from T1 to T2


I hope that helps.
Regards,
BEATRIZ

Hi Beatriz, 
 
Thank you for response.

I cant understand usage of MeshMath.
 
ı complete the shape analyses for  each group (diseased and control) with shapeAnalysisMANCOVA. Now ı have shapeAnalysisMancova's outputs for two groups. 
Which file am ı use in MeshMath to compare two groups differences? and how ı use that file ?

Hi Onur,


MeshMath vector computing comes before running shapeAnalysisMANCOVA. You have to pair your two time points and subtract them in the way we described before. That will provide with magnitude of change (distances) and vectors that show the directionality of that change from tp1 to tp2 or viceversa. 

When you have those vector files (one for each tp1-tp2 pair), those are the ones that have to be assigned a group and analyzed in shapeAnalysisMANCOVA_Wizard:
- The input file will be your vector files
- You have to specify that your input is a KWM readable file (in the tab Test Parameters).


Regards,
BEATRIZ

I am grateful to you for your help.

 I have label.hdr file for each person. All data are in ANALYSES format. Which process must ı apply to these file for prepare that file for MeshMath function?

 
I am new in image processng. for this reason ı may ask easy question , sorry for that.

Thanks ..
Onur . 

Hi Onur,

No problem, I am happy to help. Lets clarify the order of the steps you should be taking to do your longitudinal analysis:

a) ShapeAnalysisModule pipeline, that is the one you describe in http://www.nitrc.org/forum/forum.php?thread_id=2525&forum_id=1281

b) In your correspondent meshes, you apply MeshMath to compute your vectors.

c) you analyze your vector files in shapeAnalysisMANCOVA_Wizard.

You are not supposed to run MeshMath before you get your correspondent meshes with ShapeAnalysisModule, because is not even possible. Our subtraction method assumes correspondence.



Regards,
BEATRIZ

Hi Beatriz , 

 After ShapeAnalysisModule  ı got that files;

 label_para.meta
 label_surf.meta 
 label_surfSPHARM.coef
 label_surfSPHARM.meta
 label_surfSPHARM_ellalign.coef
 label_surfSPHARM_ellalign.meta
 label_surfSPHARM_procalign.meta .

 which files do ı use in MeshMath ?

It depends, if you go to http://www.nitrc.org/plugins/mwiki/index.php/spharm-pdm:Pipeline_Doc_V1.9#SPHARM you will find a description of what each one of the outputs means. para.meta and surf.meta are not interesting for your analysis since they are intermediate step outputs.

*surfSPHARM* are the correspondent, spherical harmonic representatiocn meshes. If you have pre-aligned data, my advise is that you keep your files in the original space (surfSPHARM.meta) if you want to used the SPHARM-registered ones you have two options: ellalign and procalign (see previous link for description).

You will use the selected surfSPHARM mesh for tp1 and tp2 for each individual subject, that means you will have to repeat the MeshMath command as many times as subjects you have. Makes sense?

Regards,
BEATRIZ

Hi Beatriz , 
 
 I understand that , I must follow that steps,
 My data's are , 18 controls( in 2011) , 18 controls (in 2006) and 18 diseased (in 2011) , 18 diseased(in 2006)
 
i)shapeanalysesModule pipeline for 72 data,

ii) Meshmath control1_2006.meta -subtract control1_2011.meta for 18 controls and assign them group 1.

iii) Meshmath diseased1_2006.meta -subtract diseased1_2011.meta for 18 diseased and assign them group 2.

iv) use shapeanalysisMancova group analysis for these two groups ? 

am I wrong  ?

Yes :) you are right. Just remember that the mancova will be performed over the vectors and not the meshes.
Good luck with your experiments.

BEATRIZ

Thank you for all help. I cant learn this method without you. 
 
 ı can run meshmath function. 
 
 ./MeshMath 01_surfSPHARM.meta example.vtk -subtract 02_surfSPHARM.meta

 output format is vtk. is something wrong for that ? 

Thanks ...

Onur...

You dont need the example.vtk. Your outputs will be -subtract (vectors) and -subtract.mag (distances). You will have to rename the files to txt i.e. mv ./-subtract vectors_02_01. Also be careful with the order: the vectors go from second to first. In the command above you would be computing vectors from 02 to 01.

Regards,
BEATRIZ

i) Which file will ı have to rename; -subtract , -subtract.mag ? (ı understand -subtract but ı ask for be sure)

 ii) How do ı ımport these .txt files to Mancova ? 

 my csv file is like ,

 Group,Data
 1, \home\onur\Desktop\...\example1.vtk
 0, \home\onur\Desktop\...\example1.vtk
 .
 .
 .

.
must the new csv file  be like,

Group,Data

 1, \home\onur\Desktop\...\subtract01_02.txt

 0, \home\onur\Desktop\...\example001_002.txt

 .

 .

 .

Thanks.

Yes to the two questions.
Good job!


BEATRIZ

Hi Beatriz, 

 I completed the shapeAnalysisModule. But ı cant run the ShapeAnalysisManvoca_Wizart with my csv file. 
 
 ı am sending a screenshot about about manvoca. Do u see anything wrong ? I cant understand why the Mancova_Wizard didnt run .

 Thanks... 
 ONur...

The CSV file tab is ok, but... did you set the Test parameters tab?

Regards,
BEATRIZ

Yes ı set the Test parameters tab. I send a screenshot of Test parameters tab.

And ı also tried  check "Input data will be obtained from a KWMeshVisu readable feature (vector) file instead of the 3-d coordinates in a MeshFile." .

Hi Onur,


Your study, since you use vectors and not meshes, only makes sense if you check that box (Input data KWM). I think if you check that one and still doesnt work, it might be related to your binaries. When you type in a terminal: "shapeAnalysisMANCOVA --help" do you get the help?

Regards,
BEATRIZ

>>>>When you type in a terminal: "shapeAnalysisMANCOVA --help" do you get the help?

 Yes, ı get help, now what do you recommend to me ?

 Regards,
 Onur 

Well, now we know the binary works (because the help pops out). If you run shapeAnalysisMANCOVA_Wizard do you get a file that ends in *commandline.txt? If so, you should open it, copy it in the terminal and see if you get some other error messages or if the command works.
Does it make sense?

I run the shapeAnalysisMANCOVA_Wizard but ı didint get any files.

Do you run shapeAnalysis_Wizard by just doble-clicking the binary, or do you do it by typping shapeAnalysisMANCOVA_Wizard in the terminal?
If you dont do it in the terminal, do it now in order we can get more error messages.
Thanks.

ı run the shapeAnalysisMANCOVA_Wizard in the terminal. I start the shapeAnalysiMANCOVA_Wizard by 

 ./shapeAnalysisMANCOVA_Wizard

After ı set the all parameters and click the "Run (usually it takes several minutes)" button. I got this message in terminal ;

2
fin
0numInde
0testCol
-----------------command line shapeAnalysisMANCOVA-----------------
shapeAnalysisMANCOVA /home/onur/Desktop/vec_mancova/MANCOVA.csv --hotelling --significanceLevel 0.05 --FDRdiscoveryLevel 0.05 --KWMinput --testColumn 0 --columnIndependent  --numIndependent 0 --columnGroupTypes 0 --numGroupTypes 1 --numPerms 10000 --infileColumn 1

And how about if in the terminal you type

?