Sampling Error Profile Analysis for calibration transfer in multivariate calibration

摘要: Abstract A new strategy named Sampling Error Profile Analysis (SEPA) is proposed in the optimization for some parameters piecewise direct standardization (PDS), such as number of principal components and window size, evaluation calibration transfer. Partial least squares (PLS) with mean-centering used PDS Random re-sampling carried out SEPA to obtain a series subsets build same sub-models that produce corresponding root mean square errors (RMSE), which value standard deviation are calculated. To take both accuracy stability into account, sum parameter model evaluation. The performance has been tested on two data sets: ternary mixture dataset corn dataset. Compared PDS, SEPA-PDS obtained lower prediction errors, indicating transfer would be more robust effective when using optimized by SEPA. other commonly methods slope bias correction (SBC) spectral space transformation (SST), acquired satisfactory results.