RAT Entry Functions#

The user should begin by creating a project and controls object, then the user can run RAT like shown below

    % Initialize the project class
    problem = createProject();

    % Initialize the controls class
    controls = controlsClass();

    % call the RAT function
    [problem,results] = RAT(problem,controls);

When the RAT function is called, the classes are passed into internal functions like parseClassToStructs which takes the classes and breaks them down into cells, limits, priors and more importantly converts the project class to struct.

Then, the RATMain function redirects the control flow based on what procedure is selected in controlsClass. One of the redirecting functions will call the reflectivityCalculation which starts the reflectivity calculation.

Some interesting data type changes are needed because of how things work with coder. Coder wont accept variable sized cell arrays contains variable sized arrays (strings for eg) in a field of a struct. So, look at parseClassToStructs function to understand how the data is converted.

API.RAT(project, controls)#

Runs RAT calculation for the given project and controls inputs.

Parameters:

• project (projectClass) – An instance of the projectClass.

• controls (controlsClass) – An instance of the controlsClass.

Returns:

• project (projectClass) – An instance of the projectClass with updated fit values from the calculation.

• result (struct) – The results of the calculation such as simulated reflectivities, SLD profiles etc.

Notes

For a non-Bayesian procedure, the result struct has the following fields below where nDomains, nContrasts and nParams are the the number of domains, contrasts and params respectively

Field	Type	Description
reflectivity	[nContrasts x 1] cell	calculated reflectivity values
simulation	[nContrasts x 1] cell	simulated reflectivity values
shiftedData	[nContrasts x 1] cell	Data shifted using given scale factor
backgrounds	[nContrasts x 1] cell	background values for each contrast
resolutions	[nContrasts x 1] cell	resolution values for each contrast
sldProfiles	[nContrasts x nDomains] cell	SLD profiles
layers	[nContrasts x nDomains] cell	layers parameter values for each contrast
resampledLayers	[nContrasts x nDomains] cell	resampled layer values for each contrast if resample is true otherwise an array of zeros
calculationResults	[1 x 1] struct	Chi-squared goodness of fit results. The fields are chiValues and sumChi
contrastParams	[1 x 1] struct	Parameter values for each contrast. The fields are scalefactors, bulkIn, bulkOut, subRoughs, and resample
fitParams	[1 x nParams] double	fitted parameter values
fitNames	[nParams x 1] cell	name of fitted parameters

For a Bayesian procedure, the result struct will also contain the following fields in addition to the ones above

Field	Type	Description
predictionIntervals	[1 x 1] struct	Mean, 65% and 95% confidence intervals, and chi squared goodness of fit values for the reflectivity and SLD. The fields are reflectivity, sld, and sampleChi.
confidenceIntervals	[1 x 1] struct	Confidence intervals for the Markov chain. The fields are percentile95, percentile65, and mean.
dreamParams	[1 x 1] struct	Parameters used to configure dream.
dreamOutput	[1 x 1] struct	Diagnostic output information from a DREAM run. The fields are allChains, outlierChains, runtime, iteration, modelOutput, AR, R_stat, and CR.
nestedSamplerOutput	[1 x 1] struct	Output from a nested sampler run. The fields are LogZ, LogZErr, nestSamples, and postSamples.
chain	[M x nParams] double	MCMC chains where M is the length of each chain.

API.RATMain(problemStruct, controls)#

Adds C struct names for inputs

API.parseClassToStructs(project, inputControls)#

Breaks up the classes into the relevant structures for inputting into C