This standard specifies the terms and symbols for marine detection, statistical tests for outliers, significance tests for differences between two means, validation of analytical methods, preparation and application of internal control samples, analytical quality control charts, etc. This standard applies to data processing and laboratory internal analysis quality control for seawater analysis, sediment analysis, biological pollution ecological surveys and biological detection in marine environmental monitoring. Marine atmosphere, pollutant flux surveys into the sea, marine dumping and dredged material surveys, etc. can also be used as a reference. GB 17378.2-1998 Marine Monitoring Specifications Part 2: Data Processing and Analysis Quality Control GB17378.2-1998 Standard Download Decompression Password: www.bzxz.net
This standard specifies the terms and symbols for marine detection, statistical tests for outliers, significance tests for differences between two means, validation of analytical methods, preparation and application of internal control samples, analytical quality control charts, etc. This standard applies to data processing and laboratory internal analysis quality control for seawater analysis, sediment analysis, biological pollution ecological surveys and biological detection in marine environmental monitoring. It can also be used as a reference for surveys on marine atmosphere, pollutant flux into the sea, marine dumping and dredged materials, etc.

GB 17378. 2—1998
This standard is the second part of the "Ocean Monitoring Specifications", which is revised on the basis of the HY003.2~-91 industry standard. This standard is the technical regulations for the quality control of data processing and analysis of ocean monitoring. The "Ocean Monitoring Specifications" includes the following parts: GB 17378. 1—1998
GB 17378. 2—1998
GB 17378. 3--1998
GB 17378. 4--1998
GB 17378. 5—1998
GB 17378. 6--1998
GB 17378. 7--1998
Ocean Monitoring Specifications
Ocean Monitoring Specifications
Ocean Monitoring Specifications
Ocean Monitoring Specifications
Ocean Monitoring Specifications
Ocean Monitoring Specifications
Part 1: General
Part 2: Data Processing and Analysis Quality Control Part 3: Sample Collection, Storage and Transportation Part 4: Seawater Analysis
Part 5: Sediment Analysis
Part 6: Organism Analysis
Part 7: Ecological Survey and Biological Monitoring of Offshore PollutionOcean Monitoring Specifications
Appendix A of this standard is a suggested appendix.
This standard is proposed by the State Oceanic Administration.
This standard is under the jurisdiction of the National Center for Marine Standards and Metrology. This standard was drafted by the National Marine Environmental Monitoring Center. The main drafters of this standard are: Zhang Chunming, Chen Banglong, Wang Huichang, Li Nailan, Xu Weilong. 334
1 Sample diagram
National Standard of the People's Republic of China
Specification for marine monitoring
Part 2: Data processing and quality control of analysisGB 17378.2—1998
This standard specifies the terms and symbols for marine monitoring, statistical tests for outliers, significance tests for differences between two means, validation of analytical methods, preparation and application of internal control samples, analytical quality control charts, etc. This standard is applicable to data processing and laboratory internal analysis quality control for seawater analysis, sediment analysis, organism analysis, offshore pollution ecological surveys and biological monitoring in marine environmental monitoring. Marine atmosphere, pollutant flux surveys into the sea, marine dumping and dredged material surveys, etc. can also be used for reference.
2 Referenced standards
The provisions contained in the following standards constitute the provisions of this standard through reference in this standard. When this standard is published, the versions shown are valid. All standards are subject to revision. Parties using this standard should explore the possibility of using the latest versions of the following standards. GB8170--87 Rules for Rounding Off Numerical Values
3 Definitions
This standard adopts the following definitions.
3.1 Laboratory sample laboratory sample Samples that can be sent to the laboratory for analysis and testing, divided into original samples, analytical samples, and test samples. 3.2 Raw sample
The initial sample collected on site.
3.3 Analytical sample analytical sample Samples that need to be pre-treated before entering the test. 3.4 Test sample
Samples of the substance to be tested that can be directly sent for testing. If the analytical sample does not require any treatment, the analytical sample and the test sample are consistent. 3.5 Parallel sample parallel sample
Samples taken from the same sample and independent of each other. 3.6 Standard blank standard blank
The response value of zero concentration in the standard series.
3.7 Analysis blankanalysis blank
The measurement result of the blank sample under the same conditions as the sample analysis. 3.8 Calibration curvecalibration curveApproved by the State Administration of Quality and Technical Supervision on June 22, 1998 and implemented on January 1, 1999
GB 17378. 2-1998
The correlation curve between the value of the item to be measured in the sample (X) and the signal value (Y) given by the instrument. The calibration curve is divided into standard curve and working curve.
3.9 Working curveworking curve
A calibration curve whose measurement steps of the standard series are exactly the same as those of the sample. 3.10 Standard curvestandard curve
A calibration curve whose measurement steps of the standard series are simplified compared with those of the sample. 3.11 Method sensitivitymethod sensibilityThe degree of change in the response of a method to a change in the unit concentration or unit amount of the substance to be measured. 3.12 Detection limit (X) detection limit The detection limit of an analytical method refers to the lowest concentration or amount that can be qualitatively distinguished from zero with a probability of 95%. 3.13 The lower limit of determination (X) limit of determination is the lowest concentration or amount that can be quantitatively detected as non-zero with a given probability (such as 95%). 3.14 Type I error (α error) type I error (α error) The original hypothesis is true but is rejected, also known as rejection of the truth. For example, the hypothesis that the analyte is zero is originally correct, but is mistakenly rejected and mistakenly judged as detection, and vice versa.
3.15 Type II error (β error) type I error (β error) The original hypothesis is not true, but is accepted, also known as preservation of the falsity. For example, the hypothesis that the analyte is zero is originally wrong, but is mistakenly accepted and judged as non-detection, and vice versa.
3.16 Undetection
A measurement result that is lower than the detection limit Xn.
3.17 Precision
The degree of agreement between independent test results under specified conditions. Note: Precision depends only on random errors and has nothing to do with the true value or other agreed values ​​of the measured value. 3.18 Range (R) range
The difference between the maximum and minimum values ​​in the sample.
3.19 Deviation (D) deviation
The difference between each single measurement value and the average value. 3.20 Relative deviation (RD) relative deviation The ratio of the difference between the sample component and the sample mean to the sample mean. 3.21 Standard deviation standard deviation The sum of the squares of the differences between the sample components and the sample mean divided by the square root of the sample size minus 1: S
Where: S
Standard deviation;
X,-each measured value;
X average value;
n—number of repeated measurements;
When n≥20: 8=
[1 of (xx).
(X,-x)
3.22 Relative standard deviation (RSD)
relative standard deviation
The ratio of the sample standard deviation to the sample mean. 3.23 Repeatability repeatability
GB 17378. 2 1998
The degree of consistency between independent test results under repeatability conditions. 3.24 Repeatability condition The test condition in which the same test object is independently tested in a short period of time by the same operator using the same equipment in the same laboratory.
3.25 Reproducibility
The degree of consistency between test results under reproducibility conditions. 3.26 Reproducibility conditions The test condition in which the same test object is independently tested under conditions where there are essential changes in the test laboratory, operator, test equipment, test procedure (method), and test time. 3.27 Accuracy
The degree of consistency between the test result and the true value or agreed true value of the measured object. 3.28 Error of a test
The difference between the test result and the true value (or agreed true value) of the measured object. Test errors include systematic errors and random errors. 3.29 Systematic error Systematio error In multiple measurements of the same measured object, it remains unchanged or changes according to a certain rule. Note: Systematic errors and their causes can be known or unknown. 3.30 Random error
In multiple tests on the same measured quantity, it changes in an unpredictable way due to accidental factors. Note: Random error cannot be corrected. 3.31 Statistic statistic
A function of a sample that does not depend on unknown parameters. 4 Relevant regulations
4.1 The original work records on site shall be written in hard pencil on the designated form, with correct handwriting and no erasure. When it is necessary to correct an error, draw a horizontal line on the wrong number and write the correct number above it. 4.2 Correctly record and round off the number of significant figures, record it according to the method given in 5.1, and do not arbitrarily increase or decrease the number of digits to ensure the accuracy of the data.
4.3 The original work records are important technical archives and should be classified and archived according to their preservation value. 4.4 The dimensions and number of significant figures of the test results should be reported in accordance with the specific provisions of the analysis method. If there is no such provision, the general principle is: only the last digit of a data is allowed to be an estimated (doubtful) value, and the other digits are valid (credible), and the number of significant digits of integers and decimals is determined accordingly. Due to the change of dimension, no rigid provision is made for the decimal position. For details of the calculation method, see Chapter 5. 4.5 Test results below the detection limit X~ should be reported as "not detected", but when the regional monitoring detection rate accounts for more than 1/2 (including 1/2) or less than 1/2 of the sample frequency, the undetected part can be taken as 1/2 and 1/4 of X respectively for statistical calculation. 4.6 Data generated without implementing the quality control procedures stipulated by the competent department of the business shall be regarded as doubtful data, and doubtful data shall not be used for marine environmental quality and marine environmental impact assessment. 4.7 Testing parallel samples is also one of the methods of analytical quality control. In principle, for items that are not measured simultaneously with internal control samples, double parallel analytical samples shall be tested. Only individual items such as dissolved oxygen and oil in water shall be tested in double parallel of original samples (such items do not need to be tested in double parallel of analytical samples). The relative deviation allowable value of the double seawater samples, if the original method does not have such provisions, shall be implemented according to Table 1. The relative deviation table of double sediment and organism samples can be found in GB17378.5 and GB17378.6.337
Analysis result order of magnitude
Relative deviation allowable limit, %
GB 17378. 2 -- 1998
Table 1 Relative deviation table of parallel double seawater samples
4.8 The recovery rate of spiked natural samples shall not exceed the range given by the method. If there is no such provision, refer to Table 2.
Table 2 Recovery rate allowable value table
Concentration or content range, μg/I.
Volume and weight method
Calculation 4=×100%
Recovery rate, %
60~110
80~110
90~110
95~105
4.9 In marine monitoring, if analytical methods other than those in this specification are used, method comparison and verification work must be carried out in accordance with regulations and reported to the business competent department for approval and filing.
5 Data processing
5.1 Significant figures and numerical rounding
5.1.1 Significant figures
0, 1, 2, 3, 4, 5, 6, 7, 8, 9 These numbers are called numbers, and the combination of more than one number constitutes a numerical value. The position occupied by each digit in a value is called a digit. The first digit after the decimal point is called the tenth digit, followed by the hundredth digit, thousandth digit, and the first digit before the decimal point is called the unit digit, followed by the tenth digit, hundredth digit, and dry digit. The digits on each digit in a value should be valid. Only the last digit is allowed to be an estimated digit, but its fluctuation range shall not be greater than ±1. For example, when the last digit is 5, it may be 4 or 6, and the rest of the digits are reliable digits (except for the position 0). The number of digits used to express a value is called the number of significant digits. In addition to reflecting the size of the value, the number of digits also reflects the accuracy of the value in the field of analysis. For example, 0.6705g of sodium oxalate is between 0.6 and 0.7g in terms of accuracy. The reliable digit is 0 at the thousandth digit, and the number 5 at the ten-thousandth digit is suspicious, but its fluctuation range is less than 0.0002g. The role of the digital "0" varies a lot. Whether a "0" in a value is a valid number depends on the position of the "0" and the numbers before and after it. The following four situations are common: 5.1.1.1 The "0" between non-"0" numbers, such as the three "0"s in the two numbers 2.005 and 1.025, are all valid numbers. 5.1.1.2 All "0"s after non-"0" numbers are valid numbers (except for the "0" at the end of the whole integer). For example, 2.2500 and 1.0250. 5.1.1.3 "0" that is not preceded by a non-zero number, such as the three "0"s in 0.0025, are not valid numbers and only play a positioning role. 5.1.1.4 The last "0" in an integer can be a valid number or not. For example, if you use an ordinary balance to weigh 1.5g of a reagent, if it must be expressed in mg, it should be written as 1500mg. The last two "0"s in this value are apparently valid numbers, but they are not actually, because a rough balance cannot achieve such a high degree of accuracy. In order to avoid misunderstanding, it can be expressed in exponential form. The above example can be recorded as 1.5×10°mg, or as 1500mg±100mg. This clearly shows that there are only two significant figures. 5.1.2 Original record of data
The number of significant figures of a value is the main indicator of its accuracy. In order to ensure the accuracy of the data, starting from the correct recording of the original data, any data with computational significance must be carefully estimated and the number of significant figures must be correctly recorded. For example, the minimum graduation value of a 50mL burette is 0.1mL, and because an estimated digit is allowed to be added, two decimal places can be recorded, such as 12.34mL. Recording this value shows that the 3 on the tenth place is the scale indication value, which is accurate and reliable, and the 4 on the hundredth place is an estimated interpretation, which is a suspicious number. It is known that its fluctuation range is 0.02mL, and its relative error is (0.02/12.34)×100%=0.16%. If it is only recorded as 12.3mL in the original record, it means that a relative error of 1.6% may occur. The accuracy of the data is reduced by several orders of magnitude due to the unreasonable original record. However, the number of significant figures cannot be increased arbitrarily. For example, if it is recorded as 12.340 in the previous example, it is obviously distorted because it is impossible to estimate the two digits. The number of significant figures in the original record should not be less or more. The principle of recording is to record truthfully according to the minimum scale value indicated by the instrument and meter and allow an additional estimated digit. The number of digits that can be recorded for common laboratory measuring instruments is as follows: the fourth digit after the decimal point of the one-ten-thousandth balance is the ten-thousandth place. The second digit after the decimal point of the upper III balance is the percentile. The absorbance value of the spectrophotometer is recorded to the third digit after the decimal point, which is the thousandth place. The number of significant digits recorded for glass measuring instruments must be determined based on the allowable error of the measuring instrument and the reading error. The accurate capacity of common first-class measuring instruments is recorded according to Table 3 and Table 4. Table 3
Indication of accurate capacity of first-class non-gradient pipettes Capacity indication
Capacity indication
5.1.3 Approximate calculation rules
Allowance difference
Indication of accurate capacity of first-class volumetric flasks Table 4
Allowance difference
Accurate capacity
Accurate capacity
In order to ensure that the final result contains only significant digits (with the exception of positional "0"), the following rules must be observed in the calculation: ml
5.1.3.1 The number of significant decimal places retained in the final calculation result of addition and subtraction operations should be the same as the number of decimal places with the least number of decimal places in the values ​​participating in the calculation. Example:
11.14+5.91225=17.05225→17.0511.14—5.91225-5.22775→5.23The final result of the above example can only retain two decimal places, because the last digit 4 of 11.14 itself is unreliable, and the numbers after it are even more unreliable.
The number of significant digits retained after rounding of the multiplication and division operation should be the same as the number with the least significant digits among the several values ​​participating in the operation.
GB 17378.2 1998
5.1.3.3 The number of significant digits of the logarithm of logarithm operation should be the same as the original number (true number). 5.1.3.4 The number of significant digits of the calculation result of square, cube and square root operation should be the same as the original number. 5.1.3.5 The number of significant digits of yuan, 2, 3, number
, etc. shall be determined by referring to the relevant data. 5.1.3.6 When there are more than 4 data from a normal population, the number of significant digits of its average value may be increased by one compared with the original number. 5.1.3.7 The standard deviation used to express the precision of the method or analysis result generally has only 10 digits of significant figures; when the number of measurements is large, 2 digits can be used, and at most 2 digits can be used. 5.1.3.8 The number of significant figures in the reported analysis results should be determined according to the precision of the analysis method, that is, the size of the standard deviation. Usually, the digit where the first digit of one-fourth of the standard deviation is located can be used as the mantissa of the analysis result. For example, a measurement result is 25.352, the standard deviation is 1.4, and the standard deviation of one-fourth is 0.35. The digit where the first digit is located is the tenth digit, which is the last digit of the result and can be reported as 25.4. 5.1.4 Rounding off of numerical values ​​
For details on rounding off of numerical values, please refer to the relevant provisions of GB8170-87. 5.1.4.1 Among the numbers to be discarded, if the left digit is less than 5 (excluding 5), it will be discarded, that is, the last digit to be retained will remain unchanged.
For example, 14.2432 is rounded to one decimal place: Before rounding
After rounding
5.1.4.2 Among the numbers to be discarded, if the digit on the left is greater than 5 (not including 5), then add one, that is, add one to the last digit to be retained.
For example, 26.4843 is rounded to only one decimal place: Before rounding
After rounding
5.1.4.3 Among the numbers to be discarded, if the digit on the left is equal to 5, and the digits on the right are not all "0", then add one; if the right side of 5 is all "0", if the last digit to be retained is an odd number, then add one, if it is an even number (including "0"), then do not add one. For example, the following values ​​are rounded to only one decimal place: Before rounding
After rounding|| tt||5.1.4.4 If the number to be discarded is more than two digits, it shall not be rounded off multiple times in succession. The result shall be rounded off once according to the above provisions based on the size of the ...th digit to the left of the number to be discarded. For example, 15.4546 is rounded off to an integer.
The correct way is:
Before rounding
Incorrect way:
After rounding off (result)
Before rounding off
Three rounding off
GB 17378. 2 -- 1998
First rounding
Fourth rounding (result)
Second rounding
During the rounding calculation process, the intermediate results do not need to be rounded, and the final results will be rounded to the expected number of digits. 5.2 Statistical test for outliers
The normal measurement data of a group (cluster) should come from the same population with a certain distribution; if the analysis conditions change significantly, or there is a mistake in the experimental operation, data that is significantly different from the normal data will be generated. Such data are called outliers or outliers. When it is only suspected that a certain data may distort the measurement results, but it has not been tested and determined to be an outlier, this data is called suspicious data. 5.2.1 Test of suspicious data
Removing outlier data will make the measurement results more objective; if some data with large apparent differences that are not outliers are arbitrarily deleted out of good intentions, although satisfactory data is obtained, it does not conform to objective reality. Therefore, the selection of suspicious data must be handled in accordance with the following principles.
5.2.1 .1 Carefully review and recheck the test process that produced the suspicious value. If it is a negligent error, discard it. 5.2.1.2 If no negligence is found, it should be tested according to the statistical procedure to decide whether to discard it. 5.2.2 Criteria for judging outliers
5.2.2.1 If the calculated statistic is not greater than the critical value of the significance level α=0.05, the suspicious data is normal data and should be retained. 5.2.2.2 If the calculated statistic is greater than the critical value of α=0.05 but not greater than the critical value of α=0.01 , this suspicious data is a deviant data and can be retained, and the median is taken instead of the mean value. 5.2.2.3 The calculated statistic is greater than the critical value of α-0.01. This suspicious value is an outlier and should be eliminated, and the remaining data should continue to be tested until there are no outliers in the data. 5.2.3 Test methods for outliers
Commonly used test methods are as follows.
5.2.3.1 Dixon test method
It is used for consistency test and outlier elimination test of a set of measurement data. Steps:
a. Arrange the measured values ​​repeated n times from small to large as X,, X2, XX,; b. Calculate the Q value according to the formula in Table 5;
c. According to the selected significance level α and the number of repeated measurements n, look up Table 6 to obtain the critical value Q. ; d. According to the judgment criteria of Article 5.2.2, decide whether to choose or not. If Q>Qo.01, the suspicious value is an outlier and is discarded. If Q.05The two criteria determine the selection. If Q>Qo.01, the suspicious value is an abnormal value and is discarded.The two criteria determine the selection. If Q>Qo.01, the suspicious value is an abnormal value and is discarded.0.597
The minimum value X is judged as an abnormal value and should be eliminated. GB17378.2-1998
Table 5 Dixon test statistic (Q) calculation formula n value range
8～10
11～13
The suspicious value is the minimum value X, when
X,—X
Qu=x--x
-2—X
Table 6 Dixon test critical value (Q.) table
Significance Level (α)
5.2.3.2 Grubbs test method
When the suspicious value is the maximum value X,
X,-X,-1
X,—Xi
X,X,--
Qu-XX
Qa-XX
Significance level (α)
Used for consistency test of multiple groups of measured means and test for eliminating outliers. It is also suitable for consistency test of a series of single measured values ​​in the laboratory.
Steps:
Suppose there are L groups of data, and the mean values ​​of each group are X, x, ... X. 1) Arrange the L means in order of size, record the maximum mean as Xmax, and the minimum mean as Xmin; 2) Calculate the total mean value and standard deviation s from the L means (x,): (, -)2
Wherein: x,—
——represents the mean of each group
3) Calculate the statistic ti or t2 according to the following formula based on the suspicious value max or Xmin; Xmax —x
—Xmin
4) Obtain the critical value based on the given significance level α and the number of groups L by looking up Table 7; 5) Decide on the selection based on the judgment criteria in Article 5.2.2; 6) If this method is used to test a group of data in the laboratory, change the number of groups L to the number of measurements n, and change the mean value of each group x, to the single measurement value X.
GB 17378.2—1998
For example, 10 laboratories analyzed the same sample, and their average values ​​were 4.41, 4.49, 4.30, 4.51, 4.64, 4.75, 4.81, 4.95, 5.01, and 5.39, respectively. Check whether the maximum value 5.39 is an outlier. 10
4.746 = 4.75
Xmax —
=0.30510.31
5.39—4.75
When L10, α0.05, the critical value (T.) in Table 7 is 2.18; judgment: 2.11<2.18, t0.838C>C.01, 0.92 is an outlier variance, and 0.9 is not equal to the measurement accuracy of other laboratories and should be eliminated. Table 8 Cochran maximum variance test critical value (C.) Table L
α0.05a0.01
Significance test of the difference between two means
GB 17378.2—1998
Table 8 (end)
α=0.05α=0.01
Use statistical test procedures to determine whether the difference between two groups of data is significant, so as to use data more reasonably and make correct conclusions.
Statistical theories and assumptions are recognized and adopted in the procedure to estimate the credibility of test data. In environmental analysis work, data that need to be tested for significance are often encountered, and test methods such as t test and F test are selected for this purpose. These methods have different application fields and application conditions, which are sufficient to adapt to most statistical tests of normally distributed data. Correctly processing data is an integral part of analytical quality assurance. 5.3.1 Significance test between two groups of means: t-test method The t-test method is suitable for comparative tests of two groups of data with small sample size, unknown overall variance, but equal precision. The scope and application conditions of this method are applicable to the field of environmental analysis. The t-test method has three calculation formulas to choose from, which meet different purposes and different types of statistical tests, see Table 9.
Table 9 Application fields of calculation formulas
5.3.1.2 Comparison of paired data
5.3.1.3 Comparison of two sample means
5.3.1.4 Comparison of sample mean and overall mean5.3.1.1 Criteria for t-test
Ixi-x,1—d
Sx,-X2bzxZ.net
Application fields
Comparison of different analytical methods, comparison of analytical conditions and reaction conditions, comparison of different samples before and after changes in time and space, comparison of different samples of the same method, or comparison of different methods of the same sample, comparison of storage conditions
Comparison of two-level standard substances and standard solutions, recovery rate test 345
Tip: This standard content only shows part of the intercepted content of the complete standard. If you need the complete standard, please go to the top to download the complete standard document for free.