Find Potent Parameters

The Find Potent Parameters feature varies each parameter independently, runs a simulation, and calculates the difference of the area under the curve of each plot in the current graph for that simulation. The simulations are all saved and can be viewed by clicking the Potency toggle button near the top of the Parameters Pane. To highlight a particular parameter's simulation, click the row of the parameter in the Parameters Pane. Find Potent Parameters is done on a per graph basis; so when you click "Find Potent Parameters" on a graph, it will be done for the plots on that graph only.

How to Use

This section describes a recommended way to use Find Potent Parameters.

When there are simulation curves you wish to change by changing parameter values and you're unsure which parameters to modify, Find Potent Parameters may help.

1. Right click the graph with plots of interest.
2. Click "Find Potent Parameters".
3. Wait for the algorithm to finish.
4. Click the "Potency" header of the Parameters Table, twice; this will sort descending so you can see which parameters are most effective.

Be sure all parameters are visible. The numbers in the potency columns may not be very meaningful, but their relative values carry some meaning. The larger they are, the more potent the parameter is.

To see the particular parameter's plot on the graph, click its row in the Parameter Table and its plot will be darkened a shade darker than the initial simulation (see #How Potency is Calculated for a description of initial simulation)

Zero Potency

A value of zero for the potency means a parameter did not affect any of the plots in the graph at all during the Find Potent Parameters algorithm. Or perhaps it affects the plots below the minimum representable number in a computer. This does not mean if you change the parameter it will not affect the plots at all. It just means changing the parameter by the amount the Find Potent Parameters algorithm uses will not perceptibly change the plots. Changing the parameter by a larger amount may be effective. Changing the parameter by a smaller amount may also be effective if there is a regional blip in its effectivity at a specific value. Changing the parameter in conjunction with other parameters may also be effective; or rather, changing another parameter first may then make this parameter potent.

All that said, if there are other parameters that have non-zero values, it's probably best to focus on them first.

How Potency is Calculated

1. run initial simulation and calculate area under curve of each plot
2. for each parameter, p
3. multiply p by a factor (10 as of this writing)
4. calculate the difference in the area under the curve of each plot against the initial simulation from step 1.

It may seem extreme to multiply a parameter by a factor of 10 and this may in fact not work well for some models. The idea is to have a value that generates a significant deviation from the initial simulation. In some cases this is necessary to get out of a region where not much is happening. Of course this is all subjective to a particular model and could very well be 100, 1000, or 0.01. If model simulations are cheap, the best solution here would be to sweep several orders of magnitude in both directions and display results to the user in real time. Please let us know if this is something you desire.