3dLME result

AFNI Message Board
1

Hi

I have 2 questions about my 3dLME analysis. I have pasted the code below

  1. When I look at the GLT 1 “ADHDvTD_BackvForw”, my resulting clusters are slightly different in peak and cluster size from the F Test between the same factors Diagnosis X Complexity - though the p threshold is the same. Please could someone help me understand why these are different and which one is more accurate?

  2. Is it appropriate to report the p threshold, peak and the cluster size - similar to 3dMVM?

Thank you so much for any help

Sincerely

Prerona

3dLME -prefix ${output} -jobs 5
-model ‘DxDirectionNumberAge’
-ranEff ‘~1’
-SS_type 3
-num_glt 8
-qVars ‘Age’
-gltLabel 1 ADHDvTD_BackvForw -gltCode 1 'Dx : 1ADHD -1TD Direction : -1Forward 1Backward’
-gltLabel 2 ADHDvTD_4sv3s -gltCode 2 'Dx : 1ADHD -1TD Number : -1Three 1Four’
-gltLabel 3 Age_BvF -gltCode 3 'Direction : -1Forward 1Backward Age :’
-gltLabel 4 Age_4v3 -gltCode 4 'Number : -1Three 1Four Age :’
-gltLabel 5 InADHD_BackvForw -gltCode 5 'Dx : 1ADHD Direction : -1Forward 1Backward’
-gltLabel 6 InADHD_4sv3s -gltCode 6 ‘Dx : 1ADHD Number : -1Three 1Four’
-gltLabel 7 InTD_BackvForw -gltCode 7 'Dx : 1TD Direction : -1Forward 1Backward’
-gltLabel 8 InTD_4sv3s -gltCode 8 ‘Dx : 1TD Number : -1Three 1*Four’ \

2

When I look at the GLT 1 “ADHDvTD_BackvForw”, my resulting clusters are slightly different in peak and cluster
size from the F Test between the same factors Diagnosis X Complexity - though the p threshold is the same.

For the GLT result, are you get the peak of what: effect estimate or t-stat? How much difference in terms of cluster size? Are the F-stat values equal to the t-stat squared?

Is it appropriate to report the p threshold, peak and the cluster size

That seems to be the norm in results reporting.

3

Hi

the difference in cluster volume is not big - the f test cluster size is 567 ( with peak coordinates as -34 38 -36) and for the GLT it is 586 ( with peak coordinates as -36 34 -36 )

The F-stat values are equal to the t-stat squared

"For the GLT result, are you get the peak of what: effect estimate or t-stat? " - I am not sure how to check this?

Thank you!

Prerona

4

Since I don’t know how you performed the cluster analysis, it’s difficult for me to guess the possible reasons for the cluster size difference. For example, one possibility is the sidedness for the statistic.

"For the GLT result, are you get the peak of what: effect estimate or t-stat? " - I am not sure how to check this?

Me neither because I have no idea about the command lines or procedure you implemented.

5

Thank you so much for your help!

I have run the 3dLME command as below and then done clusterize with NN1 and then I opened the cluster report window to see the list of clusters. Here the XYZ location option is set to Peak. Everything else is set to default.

3dLME -prefix ${output} -jobs 5
-model ‘DxDirectionNumberAge’
-ranEff ‘~1’
-SS_type 3
-num_glt 8
-qVars ‘Age’
-gltLabel 1 ADHDvTD_BackvForw -gltCode 1 'Dx : 1ADHD -1TD Direction : -1Forward 1Backward’
-gltLabel 2 ADHDvTD_4sv3s -gltCode 2 'Dx : 1ADHD -1TD Number : -1Three 1Four’
-gltLabel 3 Age_BvF -gltCode 3 'Direction : -1Forward 1Backward Age :’
-gltLabel 4 Age_4v3 -gltCode 4 'Number : -1Three 1Four Age :’
-gltLabel 5 InADHD_BackvForw -gltCode 5 'Dx : 1ADHD Direction : -1Forward 1Backward’
-gltLabel 6 InADHD_4sv3s -gltCode 6 ‘Dx : 1ADHD Number : -1Three 1Four’
-gltLabel 7 InTD_BackvForw -gltCode 7 'Dx : 1TD Direction : -1Forward 1Backward’
-gltLabel 8 InTD_4sv3s -gltCode 8 ‘Dx : 1TD Number : -1Three 1*Four’ \

6

If the difference in terms of cluster threshold is not big, I would not worry too much about it. The likely reason is that the F-stat was treated as one-sided. The peak is defined differently between F-stat and a GLT: the former is based on the F-stat while the latter on the effect estimate (not t-stat when you set the effect estimate as overlay).

7

Thank you!