probit_likelihood_function is the user-defined function from the file
/home/rafael/octave/probit/probit_likelihood_function.m

 - Function File: [L, P, Q, Y, THRESHOLD, SLOPE, CHISQ] =
probit_likelihood_function (X, CONTROL_VAR, BATCH_SIZE, SUCCESS_COUNT)

This function is primarily a helper function for the maximum likelihood
optimization in the `probit_analysis' function, but it can also be used
stand alone.

The input argument X is a vector of length `n', whose first `n - 1'
elements are the thresholds and the last one is the slope for the
fitted normal cumulative functions.  Input arguments CONTROL_VAR,
BATCH_SIZE, and SUCCESS_COUNT have the same meaning as in
`probit_analysis'.

Output arguments:

`L'
     Negated log likelihood computed using the fitted probabilities and
     the variables BATCH_SIZE, SUCCESS_COUNT, which determine binomial
     distributions.  (The likelohood is negated such that this function
     can be used with `fmins'.

`P'
     Cell vector containing the fitted probability values for each
     experimental condition.  Varies between 0 and 1.

`Q'
     Simply `Q{i} = 1 - P{i}'.

`Y'
     Cell vector containing the values of `SLOPE(i) * (CONTROL_VAR{i} -
     THRESHOLD)'.

`THRESHOLD'
     The first `n - 1' elements of X.

`SLOPE'
     The last element of X.

`CHISQ'
     The test statistic chi square for the fitted curves.

This function may be useful to compute values of P for plotting the
fitted curves.  In this case, drop BATCH_SIZE and SUCCESS_COUNT from
the input argument list, drop CHISQ from the output argument list, and
use a fine-grained CONTROL_VAR.

CAVEAT: Since this funcion is primarily intended for use through fmins
from `probit_analysis', no input argument checks are realized, for the
sake of efficiency.

See also: probit_analysis.