GP Likelihood Functions

GP likelihood calculation for the gaussian model

class src.magpy_rv.gp_likelihood.GPLikelihood(x, y, yerr, hparameters, kernel_name, model_y=None, model_param=None)[source]

Gaussian Process likelihood

Parameters
  • x (array or list, floats) – Time series of the radial velocity

  • y (array or list, floats) – Radial velocity values

  • yerr (array or list, floats) – Radial velocity errors

  • hparamters (dictionary) – dictionary of all hyper parameters considered

  • kernel_name (string) – name of the used kernel

  • model_y (array, floats, optional) – Array of y (rv) values from the model chosen Default to None

  • model_param (dictionary, optional) – dictionary of all model parameters considered Default to None

Raises
  • Assertion: – Raised if the hyperparameters are not a dictionary

  • Assertion: – Raised if the a model is in use and the model parameters are not a dictionary

  • KeyError: – Raised if a model is in use and model_y is not given

  • KeyError: – Raised if a model_y is given but no parameters are specified

  • KeyError: – Raised if a model_y is given but no model is in use

LogL(prior_list)[source]

Function to compute the log likelihood after applying the priors

Parameters

prior_list (list of sets of 3 objects) –

List of the priors applied. Each item in the list should countain the following 3 objects:

String of the name of the parameter the prior is applied to

String of the name of the prior

pri_create dictionary of the prior

Raises
  • Assertion: – Raised if the prior_list is not a list

  • Assertion: – Raised if the name of the parameter is not a string

  • Assertion: – Raised if the name of the prior is not a string

  • Assertion: – Raised if the prior parameters are not a dictionary

Returns

LogL – Final ln of likelihood after applying all posteriors from priors

Return type

float

compute_kernel(x1, x2)[source]
Parameters
  • x1 (array or list, floats) – array to be used in calculation of covariance matrix

  • x2 (array or list, floats) – other array to be used in the calculations of covariance matrix

Returns

covmatrix – Covariance matrix of the chosen set of varaibles x1 and x2

Return type

array, floats

logprob()[source]

Computes the natural logarith of the likelihood of the gaussian fit. Following the equation:

\[ln(L) = -\frac{n}{2} ln(2\pi) \cdot -\frac{1}{2} ln(det(K)) -\frac{1}{2} Y^T \cdot (K-1) \cdot Y\]
Returns

logL – Ln of the likelihood

Return type

float

predict(xpred, FullCov=False)[source]
Parameters
  • xpred (array, floats) – X array over which to do the prediction

  • FullCov (True/False, optional) – Want to return the full covariance? The default is False.

Returns

  • pred_mean (array, floats) – Predicted values of the y axis

  • stdev (array, floats) – Standard deviation of the y points, to be used as error

  • OR

  • np.array(pred_cov) (array, floats) – Full covariance of the data set

priors(param_name, prior_name, prior_parameters)[source]

Fuction to compute the natural logarithm of the prior likelihood

Parameters
  • hparam_name (string) – Name of the hyperparameter on which to impose the prior

  • prior_name (string) – Name of the prior to apply

  • prior_parameters (dictionary) – Dictionary containing all necessary prior parameters (get with pri_create and assign values)

Raises

Assertion: – Raised if the selected prior is not in the list of currently available priors

Returns

prior_logprob – Natural logarithm of the prior likelihood

Return type

float

residuals()[source]
Residuals internal to the computation:

RV - RV_model

Returns

res – New RVs for internal calculations

Return type

array