summary.lmcal, summary.nlscal {quantchem} | R Documentation |
A 'summary' class for 'lmcal' and 'nlscal' objects.
summary.lmcal(object, sort.models = FALSE, ...) summary.nlscal(object, sort.models = FALSE, ...)
object |
an object of class 'lmcal' or 'nlscal' |
sort.models |
should the tables be sorted by models (TRUE) or variables (FALSE). |
... |
additional arguments, currently ignored. |
The function performs summarizing of fitted calibration models and produces several tables (see below). The are printed in appropriate form, and their list is returned invisibly.
A list, consisting of following items:
coefficients |
Estimated coefficients, their standard error, significance (t) and p-value |
residuals |
Quantiles of residuals and Shapiro-Wilk test of their normality |
variances |
Quantiles of variances (without transform, with log-log, and with Box-Cox on y) ond Bartlett test for therir heteroscedascity. Calculated only, if there are at least 2 replicates for each x |
fit |
R-squared, adjusted R-squared, AIC, residual standard error, sum of squared residuals, sum of pure error and Lack-of-Fit ANOVA test |
sensitivity |
sensitivity, limit of detection and quantitation, autocorrelation of residuals, Durbin-Watson test for autocorrelation |
The p-value of Durbin-Watson statistic is *only* approximated using normal transform algorhitm! This is not critical criterion and *always* residual plot should be visually examined.
Some of values given above are not computed for 'nlscal' models.
Lukasz Komsta, with portion by Achim Zeileis (from dwtest()
)
set.seed(1234) x=rep(1:8,5) y=jitter(sqrt(x)) fit=lmcal(x,y) fit summary(fit)