Statistical functional approach for interlaboratory studies with thermal data
- 209 Downloads
A new statistical functional data analysis (FDA) approach to perform interlaboratory tests is proposed and successfully applied to thermogravimetry (TG) and differential scanning calorimetry (DSC). This functional approach prevents the typical losses of information associated to the dimension reduction processes. It allows the location and variability of the thermal curves obtained by the application of a particular test procedure. The intra- and inter-laboratory variability and location have been estimated using a FDA approach as well as the traditional reproducibility and repeatability studies. To evaluate the new approach, 105 TG curves and 90 calorimetric curves were obtained from calcium oxalate monohydrate. The obtained curves correspond to seven simulated laboratories, 15 curves per laboratory. Functional mean and variance were estimated. From a functional point of view, these descriptive statistics consider each datum as a curve or function of infinite dimension. Confidence bands were computed using smooth bootstrap resampling. A laboratory consistency study is performed in a functional context. The functional depth approach based on bootstrap resampling is a useful tool to identify outliers among the laboratories. The new FDA approach permits to identify as outliers the thermal curves obtained with old or wrong calibrations. Functional analysis of variance test based on random projections and the false discovery rate procedure (FDR) provides which laboratories obtain significant different thermal curves. This approach can be applied to perform interlaboratory test programs where the response of the test result is functional, as, for example, DSC and TG tests, without having to assume that data follow a Gaussian distribution.
KeywordsInterlaboratory study Functional data analysis Thermogravimetry Differential scanning calorimetry Laboratory consistency Data depth
This research has been partially supported by the Spanish Ministry of Science and Innovation. Grant MTM2011-22392 (ERDF included). The authors thank the two referees for very helpful comments and suggestions.
- 4.Tarrío-Saavedra J, Naya S, López-Beceiro J, Gracia-Fernández C, Artiaga R. Thermooxidative properties of biodiesels and other biological fuels. In: Montero G, Stoycheva M, editors. Biodiesel. Rijeka: INTECH; 2011. p. 47–62.Google Scholar
- 5.Practice for conducting and interlaboratory study to determine the precision of a test method. Annual Book of ASTM Standards. West Conshohocken, PA: ASTM International E691; 2004.Google Scholar
- 6.Accuracy (Trueness and precision) of measurement methods and results - Part 2: Basic method for the determination of repeatability and reproducibility of a standard measurement method. Geneva: ISO. International Standard ISO 5725-4-1994; 1994.Google Scholar
- 8.Lysiak-Pasttuszak E. Interlaboratory analytical performance studies: a way to estimate measurement uncertainty. Oceanologia. 2004;46(3):427–38.Google Scholar
- 9.Ramsay JO, Silverman BW. Functional data analysis. New York: Springer-Verlag; 2005.Google Scholar
- 15.Ramsay JO, Silverman BW. Applied functional data analysis. New York: Springer-Verlag; 2002.Google Scholar
- 18.Shen Q, Faraway JJ. An F test for linear models with functional responses. Stat Sin. 2004;14:1239–57.Google Scholar
- 20.R Development Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, 2008. http://www.R-project.org.
- 21.Febrero-Bande M, de la Fuente OM. Statistical computing in functional data analysis: the R package FDA.usc. J Stat Soft. 2012;51:1–28.Google Scholar
- 24.Miller RG. Simultaneous statistical inference. New York: Springer-Verlag; 1991.Google Scholar
- 25.Ferraty F, Vieu P. Nonparametric functional data analysis. Berlin: Springer-Verlag; 2006.Google Scholar