family_of_distributions[source]

family_of_distributions(distribution_name, get_info, *varargin)

For each of the family of distributions this script performs specific computations like number of pdf/pmf, etc.

This function contains the probability density computation, the distribution-specific fminunc to optimize the objective function, and the associated partial derivatives for each of the dependent-variable distributions that are currently covered by the toolbox (i.e., Bernoulli, normal)

Arguments:

• distribution_name: distribution name, string, e.g. 'bernoulli', 'normal'
• get_info: Cues for specific information / computation, string, e.g. 'get_nParams'
• varargin: Is either empty or has arguments depending on the computation

Returns output of all computations.

bernoulli_distribution[source]

bernoulli_distribution(get_info, input_params)

If get_info is compute_densities, computes the log probability densities of the curves specified by input_params using the bernoulli distribution. Otherwise passes parameters to fminunc_both_betas.

fminunc_bernoulli_both[source]

fminunc_bernoulli_both(betas, w, net_effects, dependent_var)

Optimizes logistic regression betas using bernoulli cost function F

Arguments:

• betas: The current betas that were used to compute likelihoods
• w: Weight vector that holds the normalized weights for P particles
• net_effects: Predictor variable Matrix (number of trials x particles)
• dependent_var: Dependent variable Matrix (number of trials x 1)

Returns:

• f: Scalar, Objective function
• g: Vector of length 2 i.e. gradients with respect to beta_0 and beta_1

normal_distribution[source]

normal_distribution(get_info, input_params)

If get_info is compute_densities, computes the log probability densities of the curves specified by input_params using the normal distribution. Otherwise passes parameters to fminunc_normal_both.

fminunc_normal_both[source]

fminunc_normal_both(betas, w, net_effects, dependent_var, dist_specific_params)

Optimizes logistic regression betas using normal cost function F

Arguments:

• betas: The current betas that were used to compute likelihoods
• w: Weight vector that holds the normalized weights for P particles
• net_effects: Predictor variable Matrix (number of trials x particles)
• dependent_var: Dependent variable Matrix (number of trials x 1)
• sigma: Used to specify variance in the Normal distribution

Returns:

• f: Scalar, Objective function
• g: Vector of length 2 i.e. gradients with respect to beta_0 and beta_1