Making a consistent custom nominal model

Custom nominal models can be very useful tools for analysing data. However inconsistencies in the models can cause a variety of issues, such as crashes or system slowdown. For these reasons it is very important to make sure that your custom nominal models are consistent, especially because often the actual cause for these problems is hard to detect.

To start off, here is a screenshot of a nominal model that is not consistent:

The inconsistency should be quite clear: the wavelengths have been specified in exact days, 70.00 and 40.00. The harmonic ratio between 70 and 40 is 7:4 and not 3:2 as has been inserted in the harmonic ratio window.

What's the best way of avoiding inconsistencies such as this? 

The easiest way to avoid this is to accept the wavelength values that ST calculates when you change the harmonic ratio. Remember that ST works with wavelengths in seconds, and so you could switch the wavelength period to seconds when checking the consistency of the nominal model, in which case this model looks like this:

So let's fix this model. The first step is to figure out where the discrepancy starts. Always start with the longest cycle and work down. Here are the longest cycles in this model:

So far so good. Let's go further down:

Again so far so good, but further down we see this:

The discrepancy is that 40.00 weeks is not half of 18.00 months. There are a few ways of automatically fixing this. We can tick the 3:1 box, then tick the 2:1 box again to recalculate a harmonic ratio of 2:1. If the cycle above also has a 2:1 ratio (which in this case it doesn't), or the one below also has a 2:1 ratio (which in this case it does) we can use the "Apply this cycle's Harmonic Ratio to others:" option. Finally we can re-select 2:1 in the Harmonic Ratio box and the Editor will recalculate.

Once we've used one of the options above the Editor then adjusts the wavelength:

38.89 weeks isn't as elegant as 40 weeks, but that is why we have the descriptor. The important point is that the wavelengths of the model must be consistent.

We then work down the model resetting the harmonic ratios for each cycle:

And finally:

We now have a consistent model: the wavelengths are perfect harmonic ratios of each other. Our labels for the cycles (the descriptors) are still neat and do refer to the cycles appropriately. While going down we will have corrected any issues that the NM had.

As a matter of interest Sentient Trader will handle some inconsistency in the nominal model. It does not have to be absolutely 100% perfect, however most analysis problems are caused by inconsistencies,so it is a good idea to get the models as consistent as possible.