HI there,
I was wondering whether AFNI has any influential diagnostics statistics that can be ran after completing a paired t-test using 3dttest++ to identify influential participants?
Thank you,
Tamara
You could use the -singletonA option for each subject versus the subject’s group.
Thank you.
Is there a specific threshold that can be used to identify whether a participant should be removed from the analysis? I am wondering whether one of my participants is highly influential. In a normal regression, I would use Cook’s D or leverage values to identify these cases, but not sure what the standard is with AFNI.
Thank you
Can you share the 3dttest++ script?
Here is the script. I ran a paired ttest between two conditions and included a covariate.
3dttest++ -prefix /imaging/Analysis/Ttest_Feb2022/FeelvsNat_Neg -paired
-setA Feel_Neg
1422 /imaging/1422/1422.results/Feel_Bad_Resp+tlrc.BRIK
1423 /imaging/1423/1423.results/Feel_Bad_Resp+tlrc.BRIK
1425 /imaging/1425/1425.results/Feel_Bad_Resp+tlrc.BRIK
1426 /imaging/1426/1426_giantmove.results/Feel_Bad_Resp+tlrc.BRIK
1428 /imaging/1428/1428.results/Feel_Bad_Resp+tlrc.BRIK
1433 /imaging/1433/1433_giantmove.results/Feel_Bad_Resp+tlrc.BRIK
1438 /imaging/1438/1438.results/Feel_Bad_Resp+tlrc.BRIK
1439 /imaging/1439/1439.results/Feel_Bad_Resp+tlrc.BRIK
1443 /imaging/1443/1443.results/Feel_Bad_Resp+tlrc.BRIK
1444 /imaging/1444/1444.results/Feel_Bad_Resp+tlrc.BRIK
1448 /imaging/1448/1448.results/Feel_Bad_Resp+tlrc.BRIK
1451 /imaging/1451/1451.results/Feel_Bad_Resp+tlrc.BRIK
1452 /imaging/1452/1452.results/Feel_Bad_Resp+tlrc.BRIK
1453 /imaging/1453/1453.results/Feel_Bad_Resp+tlrc.BRIK
1459 /imaging/1459/1459.results/Feel_Bad_Resp+tlrc.BRIK
1477 /imaging/1477/1477.results/Feel_Bad_Resp+tlrc.BRIK
1478 /imaging/1478/1478.results/Feel_Bad_Resp+tlrc.BRIK
1479 /imaging/1479/1479_giantmove.results/Feel_Bad_Resp+tlrc.BRIK
1480 /imaging/1480/1480.results/Feel_Bad_Resp+tlrc.BRIK
1486 /imaging/1486/1486.results/Feel_Bad_Resp+tlrc.BRIK
1489 /imaging/1489/1489.results/Feel_Bad_Resp+tlrc.BRIK
1490 /imaging/1490/1490.results/Feel_Bad_Resp+tlrc.BRIK
1491 /imaging/1491/1491.results/Feel_Bad_Resp+tlrc.BRIK
1431 /imaging/1431/1431.results/Feel_Bad_Resp+tlrc.BRIK
1432 /imaging/1432/1432.results/Feel_Bad_Resp+tlrc.BRIK
1437 /imaging/1437/1437.results/Feel_Bad_Resp+tlrc.BRIK
1440 /imaging/1440/1440.results/Feel_Bad_Resp+tlrc.BRIK
1441 /imaging/1441/1441.results/Feel_Bad_Resp+tlrc.BRIK
1442 /imaging/1442/1442_giantmove.results/Feel_Bad_Resp+tlrc.BRIK
1445 /imaging/1445/1445.results/Feel_Bad_Resp+tlrc.BRIK
1447 /imaging/1447/1447.results/Feel_Bad_Resp+tlrc.BRIK
1454 /imaging/1454/1454_giantmove.results/Feel_Bad_Resp+tlrc.BRIK
1455 /imaging/1455/1455.results/Feel_Bad_Resp+tlrc.BRIK
1456 /imaging/1456/1456.results/Feel_Bad_Resp+tlrc.BRIK
1460 /imaging/1460/1460.results/Feel_Bad_Resp+tlrc.BRIK
1462 /imaging/1462/1462.results/Feel_Bad_Resp+tlrc.BRIK
1463 /imaging/1463/1463.results/Feel_Bad_Resp+tlrc.BRIK
1464 /imaging/1464/1464.results/Feel_Bad_Resp+tlrc.BRIK
1465 /imaging/1465/1465.results/Feel_Bad_Resp+tlrc.BRIK
1467 /imaging/1467/1467.results/Feel_Bad_Resp+tlrc.BRIK
1468 /imaging/1468/1468.results/Feel_Bad_Resp+tlrc.BRIK
1469 /imaging/1469/1469.results/Feel_Bad_Resp+tlrc.BRIK
1471 /imaging/1471/1471_giantmove.results/Feel_Bad_Resp+tlrc.BRIK
1474 /imaging/1474/1474.results/Feel_Bad_Resp+tlrc.BRIK
1481 /imaging/1481/1481.results/Feel_Bad_Resp+tlrc.BRIK
1482 /imaging/1482/1482.results/Feel_Bad_Resp+tlrc.BRIK
1484 /imaging/1484/1484.results/Feel_Bad_Resp+tlrc.BRIK
-setB Nat_Neg
1422 /imaging/1422/1422.results/Nat_neg_Resp+tlrc.BRIK
1423 /imaging/1423/1423.results/Nat_neg_Resp+tlrc.BRIK
1425 /imaging/1425/1425.results/Nat_neg_Resp+tlrc.BRIK
1426 /imaging/1426/1426_giantmove.results/Nat_neg_Resp+tlrc.BRIK
1428 /imaging/1428/1428.results/Nat_neg_Resp+tlrc.BRIK
1433 /imaging/1433/1433_giantmove.results/Nat_neg_Resp+tlrc.BRIK
1438 /imaging/1438/1438.results/Nat_neg_Resp+tlrc.BRIK
1439 /imaging/1439/1439.results/Nat_neg_Resp+tlrc.BRIK
1443 /imaging/1443/1443.results/Nat_neg_Resp+tlrc.BRIK
1444 /imaging/1444/1444.results/Nat_neg_Resp+tlrc.BRIK
1448 /imaging/1448/1448.results/Nat_neg_Resp+tlrc.BRIK
1451 /imaging/1451/1451.results/Nat_neg_Resp+tlrc.BRIK
1452 /imaging/1452/1452.results/Nat_neg_Resp+tlrc.BRIK
1453 /imaging/1453/1453.results/Nat_neg_Resp+tlrc.BRIK
1459 /imaging/1459/1459.results/Nat_neg_Resp+tlrc.BRIK
1477 /imaging/1477/1477.results/Nat_neg_Resp+tlrc.BRIK
1478 /imaging/1478/1478.results/Nat_neg_Resp+tlrc.BRIK
1479 /imaging/1479/1479_giantmove.results/Nat_neg_Resp+tlrc.BRIK
1480 /imaging/1480/1480.results/Nat_neg_Resp+tlrc.BRIK
1486 /imaging/1486/1486.results/Nat_neg_Resp+tlrc.BRIK
1489 /imaging/1489/1489.results/Nat_neg_Resp+tlrc.BRIK
1490 /imaging/1490/1490.results/Nat_neg_Resp+tlrc.BRIK
1491 /imaging/1491/1491.results/Nat_neg_Resp+tlrc.BRIK
1431 /imaging/1431/1431.results/Nat_neg_Resp+tlrc.BRIK
1432 /imaging/1432/1432.results/Nat_neg_Resp+tlrc.BRIK
1437 /imaging/1437/1437.results/Nat_neg_Resp+tlrc.BRIK
1440 /imaging/1440/1440.results/Nat_neg_Resp+tlrc.BRIK
1441 /imaging/1441/1441.results/Nat_neg_Resp+tlrc.BRIK
1442 /imaging/1442/1442_giantmove.results/Nat_neg_Resp+tlrc.BRIK
1445 /imaging/1445/1445.results/Nat_neg_Resp+tlrc.BRIK
1447 /imaging/1447/1447.results/Nat_neg_Resp+tlrc.BRIK
1454 /imaging/1454/1454_giantmove.results/Nat_neg_Resp+tlrc.BRIK
1455 /imaging/1455/1455.results/Nat_neg_Resp+tlrc.BRIK
1456 /imaging/1456/1456.results/Nat_neg_Resp+tlrc.BRIK
1460 /imaging/1460/1460.results/Nat_neg_Resp+tlrc.BRIK
1462 /imaging/1462/1462.results/Nat_neg_Resp+tlrc.BRIK
1463 /imaging/1463/1463.results/Nat_neg_Resp+tlrc.BRIK
1464 /imaging/1464/1464.results/Nat_neg_Resp+tlrc.BRIK
1465 /imaging/1465/1465.results/Nat_neg_Resp+tlrc.BRIK
1467 /imaging/1467/1467.results/Nat_neg_Resp+tlrc.BRIK
1468 /imaging/1468/1468.results/Nat_neg_Resp+tlrc.BRIK
1469 /imaging/1469/1469.results/Nat_neg_Resp+tlrc.BRIK
1471 /imaging/1471/1471_giantmove.results/Nat_neg_Resp+tlrc.BRIK
1474 /imaging/1474/1474.results/Nat_neg_Resp+tlrc.BRIK
1481 /imaging/1481/1481.results/Nat_neg_Resp+tlrc.BRIK
1482 /imaging/1482/1482.results/Nat_neg_Resp+tlrc.BRIK
1484 /imaging/1484/1484.results/Nat_neg_Resp+tlrc.BRIK
-covariates /imaging/Analysis/TTest_Oct5_2021/Covariate/ICU_All
Removing outliers can be problematic due to the arbitrariness involved. A more principled way is to handle it through the modeling process. If you obtain the contrast between the two conditions for each subject, you can try robust regression by feeding the contrast together with the covariate to 3dMVM with its option -robust