 |
 |
Dear Alex, The design matrix for you study could be structured in the following way: For a given subject, each level of drug may be represented in a single column that includes the 6 repetitions (see columns 1 to 5 in simplified matrix below). I assume that the reversal condition is not of interest. If you wish to find areas increasingly responsive to drug level (excluding reversal), your contrast would be 4 numbers from -1 rising to +1 in equal intervals, such that the mean is zero. To find areas in which activity declines with dose, one would use the inverse contrast. You might need to consider the following confounds. 1. The first is the overall confounding effects of repeated drug administration on activity (e.g. perhaps a voxels response to the drug potentiates/habituates every time the drug/reversal is administered). This is taken care of by high pass filtering applied in SPM. 2. The second is the confounding effects of repeated administration of drug or reversal at a specific dose (e.g. perhaps potentiation/habituation is dose-specific). If we assume that this effect is linear with respect to the repetition of a specific dose, one can introduce user-specified linearly changing regressors that model this effect. Each repetition of each drug level can ascribed a value rising linearly from 1 to 6, in the same column, to represent a linear change in activity across blocks (see columns 1 to 6 in the design matrix below). Inclusions of these covariates will minimise the presence of drug dose x repetition interactions. 3. A third, more subtle potential confound may arise if you have not administered the doses in a random order (this may not have been technically possible for your study). The dose order may influence your activations over and above the value of the dose. If the drug is administered in increasing doses, there may be (perhaps cumulative) effects that explain the activity in a voxel that are different from the effects of doses administered in a random order). In this case, it is not possible to disentangle the confounding effects of dose order from dose value. I hope this is helpful. Best wishes, Narender Ramnani. Design Matrix: Ones and zeros represent blocks Drug Level Confounds 0 1 2 3 4 R 1 2 3 4 5 6 1 0 0 0 0 0 1 0 0 0 0 0 first block of no-drug 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 first block of reversal 1 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 2 1 0 0 0 0 0 3 0 0 0 0 0 0 1 0 0 0 0 0 3 0 0 0 0 0 0 1 0 0 0 0 0 3 0 0 0 0 0 0 1 0 0 0 0 0 3 0 0 0 0 0 0 1 0 0 0 0 0 3 0 0 0 0 0 0 1 0 0 0 0 0 3 1 0 0 0 0 0 4 0 0 0 0 0 0 1 0 0 0 0 0 4 0 0 0 0 0 0 1 0 0 0 0 0 4 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 4 0 0 0 0 0 0 1 0 0 0 0 0 4 1 0 0 0 0 0 5 0 0 0 0 0 0 1 0 0 0 0 0 5 0 0 0 0 0 0 1 0 0 0 0 0 5 0 0 0 0 0 0 1 0 0 0 0 0 5 0 0 0 0 0 0 1 0 0 0 0 0 5 0 0 0 0 0 0 1 0 0 0 0 0 5 1 0 0 0 0 0 6 0 0 0 0 0 last block of no-drug 0 1 0 0 0 0 0 6 0 0 0 0 0 0 1 0 0 0 0 0 6 0 0 0 0 0 0 1 0 0 0 0 0 6 0 0 0 0 0 0 1 0 0 0 0 0 6 0 0 0 0 0 0 1 0 0 0 0 0 6 last block of reversal >Now, for my current problem. I am performing a drug study. The >design is basically a simple block design with three conditions. The >subjects perform six cycles of the three conditions. This paradigm >is repeated 5 times: with no drug, low drug, medium drug, high drug, >and then again after drug reversal. I am looking for areas which are >either more or less active as a function of drug concentration. Any >suggestions on how to set up a design matrix? Should I make one big >matrix including all 5 scans? Should I use parametric modulation? >and if so, what should be the values of the regressors? > >thanks, > >Alex