Read csv-file of the variance component estimates and construct a matrix with this estimates by parts. This matrix is checked if it is positive definite or not. If this matrix is not positive definite, 2 functions may be called. If you put the paramter psOptionRatio = FALSE, makePD2() is called . For the parameter psOptionRatio = TRUE the function make_pd_rat_ev() is called. The difference is, that with make_pd_rat_ev() you have the option to whish a maximum ratio between largest and smallest eigenvalue.
The software mix99 need a paramter-File with the variance and covariance. So this function is building this paramter-file.
parameter_varCovar_mix99( ps_input_file, ps_output_file, ps_raw_output_file = NULL, pn_ratio = NULL, pn_eps = NULL, pmat_weight = NULL, pn_digits = NULL, pb_log = FALSE, plogger = NULL )
| ps_input_file | input csv-file |
|---|---|
| ps_output_file | output txt-file for the bended (co-)variances |
| ps_raw_output_file | output txt-file for the raw (co-)variances |
| pn_ratio | maximum ratio between largest and smallest eigenvalue, determines whether ratio method is used or not |
| pn_eps | lower limit of smallest eigenvalue |
| pmat_weight | weight matrix for weighted bending |
| pn_digits | number of digits to be rounded to |
| pb_log | indicator whether logs should be produced |
| plogger | log4r logger object |
if (FALSE) { # use test input csv-file sInputFile <- system.file("extdata","VCE_results.csv", package = "rvcetools") # bend matrices using Schaeffer method parameter_varCovar_mix99(ps_input_file = sInputFile, ps_output_file = 'par_varCovar_mix99_schaeffer.txt') # bend matrices using ratio method parameter_varCovar_mix99(ps_input_file = sInputFile, ps_output_file = 'par_varCovar_mix99_ratio.txt', pn_ratio = 100) # unweighted bending with minimum eigenvalue parameter_varCovar_mix99(ps_input_file = sInputFile, ps_output_file = 'par_varCovar_mix99_unweighted.txt', pn_eps = 1e-4) }