3dRegAna is a regression program that can only be used in situations without any within-subject variables.

> We would like to generate activity maps when D' is included in the model as a covariate "of interest"

For this, use 3dLME:

3dLME -prefix myOutput -jobs 8 \
-model "Dprime" \
-qVars "Dprime" \
-ranEff '~1+Dprime' \
...

> another model of brain activity when D' is included as a "nuisance" covariate.

What are the effects of interest in this case? The 4 conditions? If so, do something like (and add, for example, pairwise comparisons with -gltCode):

3dLME -prefix myOutput -jobs 8 \
-model "Dprime+condition" \
-qVars "Dprime" \
-ranEff '~1+Dprime' \
...

I assume that you want to control D' at the average among the 4 conditions. If not, use -qVarCenters to set the center value.

Gang