The Parameter Explorer¶
Press ▶ Simulate on a Model of Experiment and the Parameter Explorer opens in its own window. It compiles the MOE's equations, runs the solver, and gives you somewhere to turn the knobs.
One Parameter Explorer opens per MOE. Leave it beside the main window: the two are linked, and moving between them is the normal rhythm of work.
The window¶
Along the top: a Save button, then Edit Associations and Weights, Optimize, Inspect Simulation, and a Tolerances toggle that reveals the solver's relative and absolute error tolerances. Change a tolerance and press Enter to re-simulate with it.
On the left, the Graphs panel plots the simulation and your data. See Graphs.
On the right, the Parameters panel lists every constant in the model. This is where you explore.
Along the bottom, a status line reports the last simulation, or a progress bar with a cancel button while one is running.
The Parameters panel¶
Choosing what to show¶
Real models have hundreds of parameters, so the panel is built around filtering.
The Display buttons choose which rows appear:
| Button | Shows |
|---|---|
| Selected | Only the parameters and initial conditions of whatever is selected in the Model Realization's diagram. Click a process there and its rate-law parameters appear here. |
| Adjustable | Only parameters you have ticked as adjustable |
| All | Everything |
The Search box narrows further, and the count beside it says how many rows
survive both filters — so a search that matches nothing because of the Display
filter is not mistaken for a search that matches nothing at all.
Ctrl+F/⌘F jumps to the search box from anywhere in the panel;
Esc clears it.
A second row of buttons chooses the comparison columns: Normal, Ref
(the reference set's value), Potency (sensitivity), or Optimizer
(bounds and fitted values). The Symbols… table in the Edit Rate Law
dialog is where to look up the S…/P… symbol behind a display name.
Pop Out detaches the whole panel into its own window, which is what you want on a second monitor.
The columns¶
Four columns are always there; the comparison buttons — Normal, Ref, Potency, Optimizer — choose which of the others appear beside them.
| Column | Shown by | Meaning |
|---|---|---|
| # | always | Row number |
| Adj | always | Adjustable — tick to include this parameter in an optimization |
| Display Name | always | The parameter's readable name — a state's own name, for an initial condition |
| Value | always | The current value. Editable — this is the main thing you change. Highlighted where it differs from the reference set |
| Ref Value | Ref | The reference set's value, highlighted where the two differ |
| Potency | Potency | Its sensitivity ranking, once you have run a potency search |
| Lower, Upper | Optimizer | Search bounds for the optimizer |
| Log | Optimizer | Search this parameter on a log scale — usually right for rate constants spanning orders of magnitude |
| Best | Optimizer | The best value the last optimization found |
| FSD | Optimizer | Fractional standard deviation of the estimate — how well the data pinned this parameter down |
Edit a value and press Simulate again to see the effect. This change-and-look loop is what the window is for; the optimizer is a way of automating it once you know roughly where you are.
Parameter sets¶
The Parameter Sets toolbar holds New, Open, Open Ref, and Make Ref. See Parameter sets.
Finding the parameters that matter¶
Right-click a graph and choose Find Potent Parameters. ProcessDB multiplies each parameter by ten in turn, re-simulates, and measures how much the plotted curves move — normalised by the area under the original curves — then ranks every parameter by the total.
The result fills the Potency column. The top of that ranking is where your data has leverage; at the bottom are the parameters the plotted data barely responds to, so a fit ranging over them has little to work with. Remove Potent Parameter Plots clears the extra curves the search drew.
Inspecting a simulation numerically¶
Inspect Simulation opens a table of the raw results — every variable at every output time. Its real use is the two boxes at the top: give a Process Flux or State Value a minimum and maximum, and the table flags every point that crosses out of range.
That is the fastest way to find a concentration that has gone negative or a flux that has run away — the usual reasons a simulation fails to converge or a plot goes flat.
When a simulation fails¶
If the solver cannot complete, ProcessDB reports the error and still forwards whatever it computed to the graphs, so you can see where things went wrong. Common causes, in rough order of likelihood:
- A parameter far from anything physical — a rate constant many orders of magnitude too large makes the system stiff beyond what the solver can handle. The Inspect Simulation bounds will show you which variable blows up.
- An initial condition of zero where the model divides by it.
- A simulation window that starts before a protocol makes sense.
- Tolerances too tight for the problem; loosen them from the Tolerances toggle.
Saving¶
The Parameter Explorer has its own Save button, and it saves things the main window's Commit does not: edited parameter values and your graph layout. See Saving your work.