The one click simulate function can be used to generate a distribution of statistics as a function of a response formula that must contain a random function (like Random Normal). Here is an article about how to use the one click simulate for power analysis in Experimental Design.
For fixed effects models the formula can be as simple as: Y = 3+ 3*X + Random Normal(). If adding a random group factor to the Y model, I first create a grouping column (group, pic below) to specify row ownership to groups and another column (G) with a formula to randomly generate the effects of each group which will later be estimated as BLUPs, and the formula changes to, say: Y = 3 + 3*X + G + Random Normal(); where G refers to the group effects — see attached JMP table file for more details.
However, if the Y formula is specified as above, the simulation of BLUPs — via simulate function, as in the picture below — does not seem to evaluate G at every simulation because, I presume, the random generation of group effects does not appear directly in the Y formula; the random group effects are generated by the G column formula not by the Y column formula.
This way, if I try to simulate with Y, the only re-evaluation that is done at every simulation is that of the random error term which is in the Y formula, that is, I get a fixed effects model simulation instead. How could I then specify a formula for Y which directly generates the group effects? I guess some looping is needed but I don’t know how to do it.
I think you are correct that the issue is that the G column is not re-evaluated for each simulation... This could be remedied by including the whole formula in one column, perhaps like this:
3 + :X * 3 + If( Row() == 1 | :group != Lag( :group ), r = Random Normal(), r ) + Random Normal()
Here the variable "r" is redefined for each group. This assumes the data table is sorted by "group". The first random normal in the formula corresponds to the group effect, and the second is the residual error.
Hope this helps.
Adam
