Residual errors and model uncertainty

[evansmanu]

Dear all,

I have a quick question that has been bothering my mind. I have successfully implemented an inverse model and the results show show the following errors;
Sum of residuals (epsilons in documentation):         7.050e+00
Sum of delta/uncertainty limit:                       7.050e+00
Maximum fractional error in element concentration:    5.000e-02

I would want to understand at what level of error is the model acceptable in the case of the uncertainties above. In the case of this model, 2 model solutions determined from the simulation with the above-mentioned errors.

Regards
Evans

[dlparkhurst]

I interpret that result as 7 elements were adjusted to their maximum extent, where the extent is the uncertainty that you assigned.

Assuming you have assigned uncertainties that are reasonable to you, then any model that you find will be within the uncertainty limits. It seems reasonable that the less you have to adjust your analytical data, the better, so a value of 2 may appeal more than a value of 7 because the data have been adjusted less. Still, a model with a value of 7 is within your constrains, and so worthy of consideration.

[evansmanu]

Dear Parkhurst,
Thanks very much for your reply. Now I am clean in my mind about the issue of residual errors. I have attached an inverse model file which I implemented with the current knowledge of geology in the area under study. I want someone to have a look if I am on the right path. I followed the tutorials in the manual and the examples as well.

Regards