Statistically Coherent Calibration of X-Ray Fluorescence Spectrometry for Major Elements in Rocks and Minerals

Instituto de Enerǵıas Renovables, Universidad Nacional Autónoma de México, Temixco, Mor 62580, Mexico División de Geociencias, Instituto Potosino de Investigación en Ciencia y Tecnoloǵıa, Camino a la Presa San José # 2055, Col. Lomas 4a Sec., San Luis Potośı, SLP 78216, Mexico Instituto de Ciencias del Mar y Limnoloǵıa, Unidad de Procesos Oceánicos y Costeros, Universidad Nacional Autónoma de México, Circuito Exterior s/n, 04510 CDMX, Mexico Posgrado en Geociencias Aplicadas, Instituto Potosino de Investigación en Ciencia y Tecnoloǵıa, Camino a la Presa San José # 2055, Col. Lomas 4a Sec., San Luis Potośı, SLP 78216, Mexico Centro de Investigación en Ciencias, Instituto de Investigación en Ciencias Básicas y Aplicadas, Universidad Autónoma del Estado de Morelos, Cuernavaca, Mor 62209, Mexico Posgrado en Ingenieŕıa, Instituto de Enerǵıas Renovables, Universidad Nacional Autónoma de México, Temixco, Mor 62580, Mexico División de Materiales Avanzados, Instituto Potosino de Investigación en Ciencia y Tecnoloǵıa, Camino a la Presa San José # 2055, Col. Lomas 4a Sec., San Luis Potośı, SLP 78216, Mexico Doctorado en Ciencias, Instituto de Investigación en Ciencias Básicas y Aplicadas, Universidad Autónoma del Estado de Morelos, Cuernavaca, Mor 62209, Mexico Universidad Autónoma de Nuevo León, Facultad de Ciencias de la Tierra, Ex–Hacienda de Guadalupe, Carretera Linares–Cerro Prieto km 8, Linares, N.L. 67700, Mexico


Introduction
All modern analytical instruments require some kind of calibration of the instrumental response (y-variable) as a function of the concentration (x-variable) [1][2][3].is calibration is generally achieved through an ordinary leastsquares linear regression (OLR) model.However, such a procedure is not strictly valid because all requirements for the statistical validity of the OLR model are not fulfilled.Usually, the assumptions "independent concentration variable x is error-free or less than one-tenth of the error in the dependent response variable y" and "error in y is homoscedastic" (i.e., equal errors for all y values) are not satisfied and, therefore, more sophisticated and statistically coherent regression procedures, such as weighted leastsquares linear regression (WLR) models, should be used [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18].
To apply the WLR and compare it with the OLR, both central tendency (e.g., mean) and dispersion (e.g., confidence limits of the mean) estimates on both x-axis (concentration, generally expressed in the unit of % m/m, i.e., mass/mass unit expressed in percent) and y-axis (response, in this case XRF intensity, generally reported in the unit of kc•s −1 , i.e., kilo counts per second) variables are required.More precise (and accurate) estimates of the central tendency will also be useful for both types of regressions.erefore, precise concentrations of GRMs with the respective lowest possible "confidence limits of the mean" (referred hereafter as the "uncertainty" of the measured variable) [2,17,18] are required to apply the regression procedure.Sometimes, we had to use also the term "error" (instead of the uncertainty) because the use of the error is widespread in the literature.
We report the following five aspects: (a) evaluation of 59 GRMs to achieve the least possible uncertainties in the mean concentrations of all major elements (SiO 2 to P 2 O 5 ); (b) the comparison of regression models (OLR and WLR) applied to net drift-corrected XRF intensities before the correction of matrix effects; (c) the second (or final) comparison of both models after achieving the matrix correction as well as for the estimation of sensitivities of the regression models; (d) application of the entire procedure to four GRMs treated as "unknown" samples and their comparison with the previous literature compilations; and (e) development of a computer program to achieve the abovementioned objectives.us, the regression equations (intercept and its uncertainty, slope and its uncertainty, and linear correlation coefficient values) for each constituent from SiO 2 to P 2 O 5 and their application to similarly complex rock matrices are presented in this work.

Evaluation of Major Element Data for GRMs
A total of 59 GRMs (listed in alphabetical order in Table S1; this and four other tables are provided in Supplementary Materials), along with a procedure blank, were used in this study.
e statistical parameters obtained in these early compilations could not be directly used for instrumental calibrations due to the following reasons: (i) the statistical methods used to achieve the statistical estimates were outdated (see [17,18,48,49] for possible reasons), and the inferred statistical values were of low quality (high values of dispersion); (ii) there are still determinations reported during about 30 or more years (postcompilation years) that were not obviously available to those compilers; (iii) the precision of more recent determinations is likely to have improved due to the availability of online computers on most modern instruments; (iv) newer more reliable statistical techniques are now available for improving both precision and accuracy of the statistical inferences, e.g., the use of discordancy tests with the highest power and lowest swamping and masking effects [18,48,[50][51][52]; and (v) importantly, new computer programs have been developed by our group [52][53][54], available at http://tlaloc.ier.unam.mxfor download or online processing of data (after previous registration onto our server), which can be advantageously used for efficient processing of experimental databases.
e same kinds of objections are applicable even today for the originator's websites, such as https://gbank.gsj.jp/geostandards/welcome.html for Japanese GRMs or https:// crustal.usgs.gov/geochemical_reference_standardsfor United States GRMs.e statistical information at these websites is based on early compilations (around 30 or more years ago).Furthermore, we were unable to use the recent work [55] because this paper reported significantly larger uncertainty values as compared to those achievable from our new validated statistical procedure [51][52][53][54]; besides, updated statistical information on the mean and its uncertainty was not available in [55] for many GRMs used in our work.
e initial databases were complemented by individual data from a large number of posterior publications (∼480; Table S1), whose complete listing is available at our server http://tlaloc.ier.unam.mxunder the heading of "Quality Control."ese major element data were classified according to the analytical method groupings [56].Data from each method group were considered as a univariate statistical sample.Appropriate discordancy and significance tests were applied from thoroughly automatized software UDASys2 [52] and UDASys3 (unpublished), which, in their "recommended procedure," apply the most powerful five (two new and three conventional) recursive tests with prior application of respective single-outlier tests having nil swamping and low masking effects [48,[57][58][59][60].Although the application of discordancy tests is identical for both UDASys2 and UDASys3, the difference lies in that the latter applies the significance (ANOVA, F and t) tests in order to provide the final results automatically.
erefore, the population mean of these GRMs is now known within the narrowest possible 99% confidence limits of the mean to best represent the concentration (x) axis in the instrumental calibrations as suggested [2,5,7,10,17,18,53]. ese data (in units of % m/m; Table S2) will also be useful for those who wish to achieve instrumental calibrations or simply use them for quality control of their results for rock and mineral matrices.

XRF Instrumentation and Intensity Measurements
A wavelength dispersive X-ray fluorescence (WDXRF) spectrometer Rigaku ZSX Primus II model (rhodium X-ray tube; 4 kW maximum power) was used for this work.We made the effort to best represent the response (y) axis (x-ray intensity in the units of kilo counts per second, kc•s −1 ) for the calibrations.For each GRM, duplicate (41 samples) or even triplicate (8 samples) pressed powder pellets were prepared.First, an appropriate amount of each GRM was dried overnight in an oven at about 105 °C.For each pellet, accurately weighed 3.5 g of moisture-free GRM was thoroughly mixed with accurately weighed 3.0 g pure N,N′ethylene bis(stearamide) beads, <840 μm as wax (Sigma-Aldrich), and stored in a desiccator.Pressed powder pellets were prepared at 20 tons•inch −2 pressure (about 310 MPa).However, for 10 GRMs, sufficient material was not available; therefore, only a single pellet could be prepared but the measured intensity uncertainty (u 99 ) at the 99% confidence level was increased by a factor of 2 to take into account the sample preparation variance.Similarly, accurately weighed 6.5 g of pure N,N′-ethylene bis(stearamide) beads, <840 μm as wax, was pressed to prepare a procedure blank sample.is was done in duplicate.
For the intensity measurements, the optimum instrumental conditions were first established through preliminary experiments prior to the routine measurements (Table S2).Each pellet was run at least 8 to 10 times in a random sequence, along with two drift monitors prepared from two volcanic rocks (basalt and rhyolite) from the San Luis Potosí Volcanic Field, San Luis Potosí (central Mexico).
e peak and background measuring conditions and time periods are also listed in Table S3.Appropriate mean drift corrections from two monitors were applied to all intensity measurements.Both monitors were run randomly 8 to 10 times each day.First, the expected monitor intensity was established as an average value of the first two days when the intensities were fairly stable and reproducible.en, the average drift correction factors were calculated for each chemical element from the two monitors run in the XRF instrument periodically before and after a set of GRMs used for the calibration.ese correction factors were then applied to the bracketed GRMs for the entire period of calibration, including the first two days and analysis of "unknown" samples.Now, although the X-ray counts may obey a Poisson distribution, we are dealing with average values of count rates, which are likely to follow a normal distribution because of the central limit theorem.A normal distribution of measured intensities was also assured for each pellet from the application of discordancy tests as explained above for GRM concentrations.
e intensity results for all pellets from a given GRM were then combined, the tests applied again to the combined data, and new mean and 99% uncertainty values were calculated for X-ray intensity of each GRM. is was done to take into account the variance of the sample preparation method, which was significantly higher than the instrumental variance of intensity measurements for individual pellets.
e drift-corrected intensity values and their 99% uncertainties (kc•s −1 ) for all GRMs, along with the concentration data and their 99% uncertainties (% m/m), are listed in Table S2.

Regression Models
Two different regression models (OLR and UWLR) were used and compared in this work.e OLR model most frequently used for instrumental calibrations (x-axis concentration and y-axis response; GRM concentration and X-ray intensity, respectively, in XRF spectrometry) requires the following assumptions to be fulfilled [4,7,10,[12][13][14][15][16][17][18]: (i) all errors are in the y-axis; (ii) x-axis is either error-free or has at most 10% error of the y-axis errors; (iii) errors in both axes are normally distributed; and (iv) errors in the y-axis are homoscedastic.Some or all of these assumptions are violated in most instrumental calibrations through the OLR model.
us, from the literature on the GRMs, it has been demonstrated that the concentration axis is not error-free (see non-zero uncertainties for all GRM concentrations in Table S2) [31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][51][52][53].One can also clearly see that the errors in the intensity axis are not homoscedastic (see unequal, i.e., heteroscedastic uncertainties for any element in different GRMs in Table S2).For a heteroscedastic linear regression system, even if each error or noise term is still Gaussian, the OLR model is no longer the maximum likelihood estimate and consequently, it is no longer efficient [10].e main advantage that the WLR has over the OLR is the ability to handle regression situations in which the data points are of varying quality as is the case with most instruments including the XRF spectrometers.
However, the major disadvantage of the WLR is that the approach is based on the assumption that the weights are known exactly.ey can be estimated using several different equations or algorithms, but when the weights are produced from small numbers of replicated observations, the regression parameters can be unpredictably affected [10].In the example of the XRF calibration that we are presenting, the numbers of observations were relatively large for both the x and y axes (concentration and X-ray intensity parameters).Besides, instead of the sample variance, we used the uncertainty values (that take into account the number of observations in the formula for uncertainty or confidence limits of the mean calculations) [2,18] for estimating the weight factors.e problem of the sensitivity to outliers in the regression equations [10] was also appropriately handled by discordancy tests programmed in the UDASys and BiDASys software [53,54,61].
erefore, although frequently used, the OLR model is not statistically correct or coherent.e statistically coherent WLR, especially the uncertainty-based WLR (UWLR [17]) model, should be used.e confidence level, such as 95% or 99% (significance level of 5% or 1%, respectively, or α of 0.05 and 0.01, respectively), can be explicitly expressed in the confidence limits of the mean or uncertainty used in the UWLR model as well as to estimate the weight factors [17].We will deal with the 99% uncertainty to have the type I error small (about 1%).Unfortunately, software of most analytical instruments, including XRF spectrometers, allows only the OLR calibration.erefore, any sophisticated regression model, such as the UWLR, will have to be applied outside the instrumental software.
us, the probability concept (99% confidence level) can be explicitly used in the UWLR model for weight factors based on the inverse of the squared 99% uncertainty of the mean.

Ordinary Least-Squares Linear Regression (OLR) Model.
Let us assume that we have a series of n reference materials or standard calibrators having individual mean concentrations x i with respective uncertainties u x i where i varies from 1 to n.In order to calibrate an instrument, each of these n calibrators were run several times, obtaining individual mean responses y i with respective uncertainties u y i where i varies from 1 to n. us, we have n bivariate concentrationresponse data pairs or calibrators (x i , y i ) with the respective uncertainties (u x i , u y i ).
We can apply the OLR model to these data for obtaining a calibration equation.e OLR fits a least-squares linear equation to the n pairs (x i , y i ) but does not take into account the respective uncertainties (u x i , u y i ).
e general regression equation for the OLR is as follows (the subscript O is for the OLR model): where m is the slope, u m is the resulting uncertainty in the slope, b is the intercept, u b is the resulting uncertainty in the intercept, x is the independent variable, y O is the dependent variable from the OLR model, and u y O is the resulting uncertainty in y. e following equations allow the calculations of these parameters: where x and y are, respectively, the mean values of the x and y variables: where  y i is the value of y O for x i in equation ( 1) and t is the Student's t test value for (n − 2) degrees of freedom, and the superscript CL is the confidence level, generally 95% or 99%: It is a general practice in most instrumental calibrations to ignore all uncertainties in equation ( 1) and use an OLR equation without any error (or uncertainty) as follows: e resulting standard deviation values of repeat measurements of unknown samples are reported as the final errors.However, these are only partial errors because the errors in the calibration equation ( 1) are not taken into account.In this work, we will use equation (1) to report total errors (in fact, 99% uncertainties) for the OLR model.

Uncertainty Weighted Least-Squares Linear Regression (UWLR) Model.
For the UWLR model, the n pairs (x i , y i ) of calibrators as well as the respective uncertainties (u x i , u y i ) are taken into account in order to achieve the best leastsquares linear fit.
e uncertainties u x i in the x-axis are first propagated to the y-axis, combined with the u y i , and the total uncertainty u i values on the y-axis are used for the weighting factors [2,10,17,18,61]: e weights are calculated from u i as follows: where w i values have the following property: us, the UWLR fits a linear equation to the n pairs (x i , y i ) with the respective weighting factors w i as follows (the subscript UW is for the UWLR model): Note that this regression line will pass closer to the data with lesser uncertainty u i .e intercept and slope variables and their uncertainties are calculated from the following equations: where x UW and y UW are, respectively, the weighted mean values of the x and y variables: where  y iUW is the value of y UW for x i in equation ( 9): e best regression equation for a calibration curve should have the following characteristics (without distinguishing the subscripts O and UW): (i) intercept b small approaching to zero; (ii) slope m large; and (iii) both u b and u m small.Further, the quality of the regression, whether a calibration curve or any other bivariate relationship, is also expressed as the linear regression coefficient (r; r O and r UW , respectively, for the OLR and UWLR), which is ideally +1.00000 for a calibration curve [5,18,61].

Original Drift-Corrected Net Intensities and GRM Concentrations: e First Set of Two Regression Equations for Each
Element. e evaluations for both regression types on the drift-corrected net intensity-concentration (Int-Conc) relationships (Table S2) for all major elements from SiO 2 to P 2 O 5 were performed (Table S4), for which the new online software BiDASys was used [61] at http: //tlaloc.ier.unam.mx.BiDASys allows the application of the conventional OLR as well as the newly proposed UWLR model [17] and provides the output of all regression parameters in an Excel ® file.Contrary to the common practice, we will refrain from showing the numerous x-y (variable x is drift-corrected net intensity "Int" and variable y is the GRM concentration "Conc") plots. is is because Table S4 statistically quantifies the visual interpretation of such diagrams.e quality parameters (standard errors se b and se m , uncertainty u b and u m , and linear correlation coefficient r and its squared value R 2 parameters) are reported in Table S4.Because we are using these several different quality parameters, the concern against the use of solely R 2 parameter [62] is not important for comparison purposes.
We will explain the implications of the statistical results for the first element SiO 2 ; the statistics for other elements (Table S4) can be similarly understood.e OLR regression equation from the first row of statistical information in Table S4 is as follows (after the element SiO 2 , subscript O is for the OLR and p is for provisional concentration; note many decimal places are used for the regression variables in such equations, because these values are not final results, and we should not introduce rounding errors during the calculation stage): � 11.47071( ± 4.90411) Similarly, the UWLR equation from the second row of statistical information in Table S4 is as follows: e implications of these regression equations can be understood from the comparison of the uncertainties of the intercept and slope, which are lower for the UWLR (equation ( 14)) than for the OLR (equation ( 13)). is means that the uncertainty of the calculated concentration will be lower for the UWLR than for the OLR.Correspondingly, the r value for the UWLR (0.99004, n � 60; R 2 � 0.98017) is much higher than that for the OLR (0.95229, n � 60; R 2 � 0.90687; Table S4).Similar trend in the r (and R 2 ) values was obtained for all other elements except MnO (Table S4).

Matrix-Effect-Corrected Intensities and GRM Concentrations: e Second Set of Two Regression Equations for Each
Element.Matrix correction is certainly required because the abovementioned least-squares linear regression fits are far from "perfect" (r ≠ +1.00000; in fact, r < 1; n � 60; r � 0.95229-0.99638for the OLR and r � 0.97715-0.99760for the UWLR; Table S4).ere is a vast literature on the subject of matrix effects in XRF and their correction procedures [63][64][65][66][67][68][69][70][71][72][73][74][75].In this study, the Lachance-Traill algorithm [73] was used for the matrix effect correction [63,71]. is was done outside the XRF instrument software.In a review of the existing algorithms, Rousseau [63] showed that the Lachance-Traill algorithm could be considered as one of the most appropriate procedures for the matrix effect correction because other algorithms have limited application range or lack of accuracy.us, for each element from SiO 2 to P 2 O 5 , a system of overdetermined equations was solved and the resulting alpha coefficients were used to correct all intensities for matrix effects.
From the alpha coefficients, matrix-corrected intensities and improved concentration values for the GRMs and their uncertainties were calculated iteratively under the condition that the convergence parameter (absolute relative difference Journal of Spectroscopy of the GRM calculated and input concentrations) for each compositional constituent (SiO 2 to P 2 O 5 ) be minimized.
New regression equations for achieving the corrected concentrations were established from the relationship of the calculated GRM concentrations (ConcCalc) and the original GRM concentrations (Conc) given in Table S2, for which the online BiDASys software [61] was used at http://tlaloc.ier.unam.mx.ese equations can be formulated from the regression coefficient values given in Table S4 (see ConcCalc-Conc rows corresponding to the OLR and UWLR).Again, we will highlight their significance for SiO 2 only.
e OLR regression equation from the third row of statistical information in Table S4 is as follows: where the subscripts O and c stand for the OLR model and calculated concentration (ConcCalc), respectively.Similarly, the UWLR equation from the fourth row of statistical information in Table S4 is as follows: where the subscripts UW and c stand for the UWLR model and calculated concentration (ConcCalc), respectively.Equations ( 15) and (16) show that the concentration values from the UWLR would be more reliable (lesser uncertainty values in both intercept and slope) than the OLR model.e r value is higher for the UWLR (0.99704, n � 60; R 2 � 0.99408; Table S4) than the OLR (0.97710, n � 60; R 2 � 0.95472).
After the matrix correction, in fact most regression equations are better because all r and R 2 values are higher for both OLR and UWLR than without the correction (Table S4; Figure 1 for r only).For the OLR, the matrix correction increased the r values (n � 60) from 0.95229-0.99638(R 2 � 0.90687 − 0.99277) to 0.97710-0.99992(R 2 � 0.95472 − 0.99994).Similarly, for the UWLR, this increase was from 0.97715-0.99760(R 2 � 0.95472 − 0.99521) to 0.99704-0.99993(R 2 � 0.99408 − 0.99986).us, after matrix correction, all r values increased for both OLR and UWLR.For the UWLR, the r values approached the ideal value of +1.00000 (Figure 1).One has to keep in mind that when the r values are closer to the maximum possible value of 1 (the "ideal" fit), the improvement expressed by the actual (absolute) value of r will apparently be small.However, as long as the r value increases for the UWLR as compared to the OLR (Figure 1; Table S4), we can objectively infer that the UWLR is a better regression model than the OLR.
Before the matrix correction, the intercepts of the Int-Conc regression lines were closer to zero for the UWLR (range ∼−0.013 to +0.011) than for the OLR (range ∼−2.098 to +11.47) model (Table S4; Figure 2).e same is true for the intercept values (ConcCalc-Conc relationship) after the matrix correction (∼−0.025 to +0.021 for the UWLR and ∼−0.110 to +1.87 for the OLR).
Finally, the uncertainties on both intercept and slope parameters were mostly lower for the UWLR than the OLR (Table S4).We highlight these differences (lower uncertainties for the UWLR) from dimensionless (free of the measurement units) parameters δu b and δu m defined as follows: Plots of these two parameters are presented in Figure 3.If u b O > u b UW , the δu b will be positive, otherwise it will be negative.e same is true for δu m .For the comparison of two models OLR and UWLR before the matrix correction, the uncertainty for the UWLR were lower than the OLR for 7 elements (positive δu b and δu m ), whereas for after the matrix correction, it was so for 8 elements (out of 10; Figure 3).e exceptions were for 3 elements Mno, CaO, and P 2 O 5 (negative δu b and δu m ) for the uncertainties before matrix

6
Journal of Spectroscopy correction and for 2 elements MnO and MgO for those after the matrix correction (Figure 3).Even for the exceptions of the elements MnO and MgO, the UWLR values should be usable (Table S4), i.e., it is not actually necessary to resort to the OLR model for these two exceptions (2 out of 10 cases).us, we can use the UWLR model for all purposes.

Sensitivities of Major Elements
We calculated the sensitivities as the slope of the Conc-IntCorr (GRM concentrations of Table S2 and matrixcorrected intensities of Table S5; see Supplementary Materials at http://tlaloc.ier.unam.mx)from the regression curve (line) for all 10 elements and for both models (Table 1).
Because the r values are significantly high (all >0.961, n � 60; Table 1) and the residuals are randomly distributed (graphs not shown), the straight line is the most likely, statistically valid fit for the concentration-matrix-corrected intensity data [5,17,18].erefore, the slope of the regression line represents an average sensitivity value for a given element under the chosen working conditions (Table S3).e intercept values were closer to zero (zero being the theoretically ideal intercept) for the UWLR regression (∼−0.113 to +0.104; Table 1) as compared to the OLR (∼−47.8 to +12.3; Table 1).e sensitivity values represented by the slopes of the regression lines (Table 1) were generally similar for both models (∼2.69-46.17kc•s −1 •% −1 for the OLR and ∼2.78-59.69kc•s −1 •% −1 for the UWLR).e sensitivity actually depends on the measuring conditions (Table S3), which were the same for both models.
For the matrix-corrected intensity-concentration (IntCorr-Conc) regressions, the parameters are listed in Table 2.All intercepts for the UWLR model, without exception, were closer to zero as compared to the OLR model.
is confirms the superiority of the UWLR model.

Application to Rock Matrices
e calibrations achieved in this work (Table S4) were applied to the analysis of four GRMs (attapulgite or Fuller's earth clay ATT1; bentonite clay CSB1; granite GH; and tonalite TLM1) taken as "unknown" samples.ese GRMs, having similarly complex matrices as the calibration samples, were not included in the calibrations because their mean values were available only from early description or compilations (for ATT1 and CSB1 [76]; for GH [77]; and for TLM1 [78]).We were unsuccessful in complementing these "old" concentration values with newer ones for these GRMs.
erefore, these GRMs were used as unknown samples.ey were analysed in exactly the same manner as the calibration samples.
All calculations for the unknown samples were done outside the instrumental software.e drift-corrected net intensities and the corresponding uncertainties were processed from the first set of two regression equations (Int-Conc OLR and UWLR models; Table S4) to obtain provisional concentration and uncertainty values.e horizontal line at the intercept value of zero represents the "ideal" intercept.
provisional concentrations were then used to obtain matrix corrections for each sample.
e method was iteratively applied with the newer concentrations to obtain the final calculated concentration values (Table 3).ese calculated concentration values were used to compute the final mean concentrations (x) and 99% uncertainties of the mean (u 99 ) for each sample from the second sets of regression equations (ConcCalc-Conc, OLR and UWLR models; Table S4).
e loss on ignition (LOI) was required to optimise the final results.e results are listed in Table 3 and compared with the literature compilations [75][76][77].On the other hand, because 99% uncertainties were not reported in the original compilations, they were computed for the comparison from the standard deviation, number of determinations, and appropriate two-sided t values at 99% confidence level [2,18].
Firstly, although the mean concentration values determined by the OLR and UWLR models showed a general agreement, the 99% uncertainty values (u 99 ; Table 3) were generally lower for the UWLR models, which clearly ) showed some differences with the preliminary values obtained by the originators of these GRMs [75].ese values for comparison were obtained in only one laboratory.e errors (uncertainties) reported in the literature were underestimated, because they did not include those resulting from the calibrations.Furthermore, the accuracy data of the originator's laboratory were not reported [75], such as the results for established GRMs and their comparison to other laboratories.

Computer Program XRFCalcUnknown
An online computer program JSpectrom_XRFCalcUnknown will be available at our server https://tlaloc.ier.unam.mxfor use for unknown samples, which will guide other users to achieve the UWLR calibration outside of the instrumental software and its routine application to unknown samples. is program incorporates the iteration process to achieve reliable  A novel aspect of the present work is that total 99% uncertainty can be calculated for individual datum in a given sample (treated as unknown; Table 3). is innovation if put into practice can entirely change the geochemical literature, and in fact make geochemistry a more quantitative science.Further, if an appropriate GRM is analysed as unknown and the analytical data (both mean and total uncertainly) are reported along with the field samples, the data accuracy can be statistically judged from such reports.

Conclusions
e XRF spectrometer calibrated under both the OLR and UWLR models clearly showed that the UWLR provides more reliable results (lower uncertainty estimates) than the OLR model commonly practiced for most XRF instruments.e sensitivity and LOD values presented for both models also supported the use of the UWLR model.
e UWLR model should therefore be used routinely in such calibrations.e use of a large number of wellcharacterized GRMs is also recommended for this purpose as illustrated in the present work.e application of our procedure was well documented for the analysis of similarly complex rock matrices.e reporting of total uncertainty values for individual datum is highly recommended for all future geochemical research.is work for the XRF shows that such a practice is easy to achieve in any other analytical calibration procedures.As the major conclusion, we can confirm that the statistically coherent WLR model was shown to perform better than the frequently used conventional statistically incoherent OLR model.

TiO 2 Fe 2 O 3 tFigure 1 :
Figure 1: Linear correlation coefficient (r) values for the ordinary least-squares linear regression (OLR) and uncertainty-based weighted least-squares linear regression (UWLR) models for the XRF calibration of major elements (SiO 2 to P 2 O 5 ) in rocks and minerals.OLR1: OLR model 1 for Int-Conc before matrix correction; UWLR2: UWLR model 2 for Int-Conc before matrix correction; OLR: OLR model 3 for ConcCalc-Conc after matrix correction; and UWLR4: UWLR model 4 for ConcCalc-Conc after matrix correction.Symbols are shown as inset.e horizontal line at the r value of 1 represents the "ideal" or "perfect" linear fit. e

Figure 3 :
Figure 3: New parameters δu b (for the intercept) and δu m (for the slope) for the evaluation of intercept (b) and slope (m) of two regression models (OLR: ordinary least-squares linear regression and UWLR: uncertainty-based weighted least-squares linear regression) before (OUW12) and after (OUW34) the matrix correction.e horizontal solid line at the y value of zero represents the line with no difference in the uncertainties of the two models.e arrows indicate that these data plotted above the scale are used for the diagram.

Figure 2 :
Figure 2: Intercept (b) values for the ordinary least-squares linear regression (OLR) and uncertainty-based weighted least-squares linear regression (UWLR) models for the XRF calibration of major elements (SiO 2 to P 2 O 5 ) in rocks and minerals.Symbols are shown as inset.For abbreviations, see Figure 1.Note some intercept values plotted outside the graph; this is indicated by arrows next to the data point.ehorizontal line at the intercept value of zero represents the "ideal" intercept.