Final Tables of Usables Ns
These numbers reflect subjects that pass QC for each task. They reflect the largest possible sample, per task, for analysis, although additional subjects may be excluded by each investigator.
The top table, under both A and B, shows the number of subjects which were completely excluded ("N excl") from either A or B scans--this includes subjects who were withdrawn from the overall study (Completed Status = Patient Withdrew and 5C_Status = Patient Withdrawn/LA2K Unusable) OR from the LA5C component (LA5C Unusable). The latter includes incomplete scans, scan discontinue, artifacts rendering all data unusable, etc.; essentially cases where all data collected from a given scan are unusable.
The bottom table, under both A and B, shows the breakdown of subjects in each Group, and breakdown of subjects on each Scanner within each Group. For example, under A scans, under BART-raw, there are 123 total controls with usable data (after excluding for reasons above and after QC); of these 123 controls, 101 were scanned at BMC and 22 were scanned at CCN.
The difference in "N not excl" in the top table and "N share", for each task, in the bottom table (for both A and B) reflect the number of subjects for each task that are not flagged for distribution following QC. This is a task-specific number. For example, there are 132 Controls that are not excluded from the total set of A scans; however, under BART-raw, there are 123 Controls to be shared. This means that 9 subjects were further excluded following QC of BART data. Furthermore, under BART-1st levels Model 2, there are 118 Controls to be shared. This means that a further 5 were excluded following QC of this BART model. An exclusion following QC, at the level of raw data, may be because of excessive motion during that task, a problem with the high-resolution anatomical image that prevents registration, or an artifact specific to that scan; an exclusion following QC, at the level of first-level data, may be because of poor task performance which renders the model rank-deficient.
For reporting of Ns, you do need to figure out the N excluded and N usable on a task-by-task basis. Here are the generic exclusion numbers (those that provided some scan data but are completely excluded/unusable):
"A portion of the total sample took part in two separate fMRI sessions, which each included one-hour of behavioral testing and a one-hour scan on the same day (order counterbalanced across subjects). 139 healthy adults completed at least a portion of the scan sessions; of those, 7 were excluded altogether (4 for missing an MPRAGE scan, 2 for withdrawing from the study, and 1 for being disqualified)".
"A portion of the total sample took part in two separate fMRI sessions, which each included one-hour of behavioral testing and a one-hour scan on the same day (order counterbalanced across subjects). 294 total participants completed at least a portion of the scan sessions; of those, 16 were excluded altogether (9 for missing an MPRAGE scan, 3 for poor quality data, 2 for withdrawing from the study, and 2 for being disqualified)".
Then the task-specific excluded N needs to be determined (example provided below):
"In addition, 14 were excluded from the present BART analyses (6 for excessive motion, 6 for poor task performance, 1 for an unusable MPARGE, and 1 for an incomplete BART scan)."
Link back to LA5C page.