Tips for exploring maps that are difficult to find a consistent optimal solution for.
Maps may be difficult to optimize or unstable for a variety of reasons, a common one with larger maps being simply that it is difficult to find a global optima and so many different local optima are found each time.
One approach that can sometimes
help is to consider running the optimizer with options = list(dim_annealing = TRUE)
(see see vignette("intro-to-antigenic-cartography") for an explanation of the
dimensional annealing approach). However be wary that in our experience, while applying
dimensional annealing can sometimes significantly speed up finding a better minima, it
can also sometimes be more prone to getting stuck in worse local optima.
If there are many missing or non-detectable titers it is also
possible that points in map are too poorly connected to find a robust
solution, to check this see mapCohesion().
Other map diagnostic functions:
agCohesion(),
bootstrapBlobs(),
bootstrapMap(),
checkHemisphering(),
dimensionTestMap(),
logtiterTable(),
map-table-distances,
mapBootstrapCoords,
mapDistances(),
mapRelaxed(),
mapResiduals(),
pointStress,
ptBootstrapBlob,
ptBootstrapCoords(),
ptLeverage,
ptTriangulationBlob,
recalculateStress(),
stressTable(),
tableColbases(),
tableDistances(),
triangulationBlobs()