Solar Notch-Delay Model Released

First, my sincere congratulations to David Evans and Joanne Nova for releasing a portion of their work on their “notch filter” model. It is very good news to see that they have done so. It obviously represents a huge amount of work on their part.
Unfortunately, I was unable to get it to run. I followed the instructions, up to the following point:

Suppose you want to run the solar model with, say, the P25 set of parameters (recommended).
1. Go to the “Comparisons” sheet and and locate the “Save and Recall Scenario” area. ” From the drop down list there, choose something like “Mix of CO2 and solar, from 1770” or “Nearly all solar, from 1900 (reconstructed TSI)”, then click “Recall”. This puts default parameters into the solar model (on the “Analysis” sheet) and the other models (on the “Models” sheet), which you can overwrite as you see fit (only write numbers into the cells with white backgrounds). It also puts appropriate parameters in the “Comparisons” sheet for running that scenario.
2. On the “Analysis” sheet, locate the “Save and Recall Solar Model Settings” area, choose “P25” from the drop down list, and press the “Recall” button. That loads the model parameters intot he solar model, and draws the transfer function and step response on the “Analysis” sheet.3.
On the “Comparisons” sheet, press “Compute All”. The big graph appears below the “Comparisons” settings, after a few seconds.

At that point I got the following error:

Run-time error ’53
File not found: kernel32

Searching the VBA code finds the following two lines in the module “CTimer”, which are the only two lines referencing the model:

Private Declare Function QueryPerformanceCounter Lib “kernel32” (lpPerformanceCount As LARGE_INTEGER) As Long
Private Declare Function QueryPerformanceFrequency Lib “kernel32” (lpFrequency As LARGE_INTEGER) As Long

It appears that these are part of a timer which keeps track of how long it takes the program to compute the answer. I suspect that the difference may be platform-dependent, as I’m working on a Mac. However, those are just guesses. In any case, I’m not able to run the model.
Finally, as Leif pointed out above, there is a deeper problem. They have only published a portion of their work. On the sheet “Storage”, they list the arbitrary parameters for their model. There are nine tunable parameters, and two binary (true/false) parameters, for a total of 11 parameters.
I must point out again the very cogent comment about arbitrary parameters by John von Neumann as related by Freeman Dyson, viz:

In desperation I asked Fermi whether
he was not impressed by the agreement
between our calculated numbers and his
measured numbers. He replied, “How
many arbitrary parameters did you use
for your calculations?” I thought for a
moment about our cut-off procedures
and said, “Four.” He said, “I remember
my friend Johnny von Neumann used to
say, with four parameters I can fit an
elephant, and with five I can make him
wiggle his trunk.”With that, the conversation
was over. I thanked Fermi for his
time and trouble,and sadly took the next
bus back to Ithaca to tell the bad news
to the students.

I shudder to think what Enrico Fermi would have said about a model with 11 parameters … regardless, I urge readers to keep in mind that obtaining a good fit from a model with 11 arbitrary parameters is meaningless.
In any case, they list a total of 25 different parameter sets. It is not clear why there are so many sets of parameters. The problem is that they have not shown how those parameters were obtained. This makes it impossible to do “out of sample” testing with their method.
Second, I had previously commented over at the JoNova blog that they should have done “out of sample” testing of their choice of parameters. Jo assured me that such tests had indeed been done.
However, they have not released those tests either.
As a result, even if I could run the model, it is impossible to do the most basic of tests of the model, the out-of-sample test.
Please be clear that I am NOT saying that David and Jo have evinced bad faith in this. From everything I’ve seen, they are honest people working to understand a most complex system, and they have put a huge amount of time and effort into the project.
However, even at this late date they have not revealed the information needed to do even the simplest testing of their model, nor even the results of the out-of-sample testing that Jo said is already done.
As a result, sadly, their model remains an advertisement, and not science in any form.
w.