alt_5pl_model.Rdnls.multstart::nls_multstart is used to calculate a 5 parameter logistic regression. Multiparametric regression requires to guess starting values. This guessing is done based on standard values (for aa, dd) and on a previous regression made by the bioplex biomanager (for cc, dd, gg). Sometimes, results of model fitting have not been consistent. Try to vary n_iter or simply repeat the fitting.
alt_5pl_model( list, n_iter = 1000, weights = c("lower", "upper", "none"), model_name = NULL, exag_factor = 10, start_pars = "std_curve_pars", ... )
| list | one list entry as derived from plexr::read_bioplex |
|---|---|
| n_iter | number of iterations for nls.multstart::nls_multstart |
| weights | on which end of the regression does good fitting to standards is required most (e.g. where do the measured values lie), lower range or upper range or no weights (none); alternatively provide a numeric vector of weights - has to be same length as list$standard$FI; rows of list$standard will be order increasingly by FI before fitting; take this into account when providing numeric weights. |
| model_name | if NULL, just the model is returned; if model_name is not NULL the model is added under that name to list and list is returned |
| exag_factor | exaggeration factor used to increase range of allowed range of gg, bb, cc from std_curve_pars |
| start_pars | name of list entry that contains start_parameters for nls; ignored if not found or if NULL |
| ... | additional parameters passed to nls.multstart::nls_multstart; e.g. or supp_errors = "Y" |
plain nls_model when is.null(model_name); or appended list with nls_model named model_name and start_pars
Model for fitting is FI ~ dd + (aa - dd) / ((1 + (Conc / cc)^bb))^gg. dd corresponds to the horizontal asymptote for x -> -Inf, aa corresponds to the horizontal asymptote for x -> Inf. (https://laustep.github.io/stlahblog/posts/5pl.html)