|Description:|| Over the last few years the MeqTrees software system that is being developed at ASTRON, with collaboration from DRAO, has emerged as a rather promising simulation/calibration package. An important node within the MeqTrees system is the Condeq node, which generates condition equations for the Solver node. Among other things, a Condeq node compares the values of observed data with values that are predicted on the basis of a (parametrized) model. The resulting differences are used to solve for better values of (subsets of) model parameters. |
As the solution converges, we expect the differences between the observed and the predicted values to decrease to zero. However, this does not always happen, for instance if we use an incomplete instrumental model, or if we try to solve for the wrong subset of model parameters. The MeqTrees system offers very extended visualization, which allows the user to detect whether something is wrong, and helps to find a way out of the problem.
In this particular case, the stubbornly non-zero residuals take the shape of a smiling (smirking?) face, which seems to mock the user. Those who still regard data reduction software as a solicitous and efficient servant will find it hard to take, being jeered at in this way. But most of us are resigned to the fact that the servant is a callous knave, who mercilessly amplifies our lack of understanding. Our highly trained scientific minds react to this outrage with a mixture of humility, superstition, appeasement, mood-swings, human (student) sacrifice, self-immolation and finally, inevitably, the Stockholm syndrome.