Statistics in spectrometry
In the spectrometer, the radiation of the analytes is converted into electrical currents by CCD and stored capacitors. The resulting voltages are proportional to the concentrations of the analytes in the sample. These measured quantities are used alone, or in a ration to another quantity as a measure of the concentration. The measured value may be in the analogue or digital form.
Analogue readings are characterized by continuous deflection in front of a scale. The observer compares deflection and the scale, and converts reading into figures. In so doing he must interpolate. The number of digits which can be read- off is determined by the fineness of the scale and his ability to estimate. A reliable estimate can be made down to 1/5 of a scale mark interval. Reading are given in “scale graduations”.
Digital reading are characterized by conversion of the analogue value into figures by means of an analogue value into figures by means of an digital converter. Readings are not continuous, and interpolation is not possible. Reading accuracy is determined by the number digits built into the equipment: in addition- again connected with the design- the last digit is unreliable. A common accuracy value is ±1 unit of the last digit. The reading is given in digits.
Each measurement contains concealed errors caused by sampling and process parameters not being held constant during measurements. Errors may be systematic and random.
In the case of spectrochemical analyses, systematic errors due to sampling techniques are the most frequent. In addition, there are systematic excitation errors eg. with SDAR, caused by systematic errors in sampling techniques.
As a typical example, if magnetics chucks for grinding, they become permanent magnets. Since the discharge current establishes its own magnetic field, the discharge channel, depending on the position of sample, is deflected in various directions. This can lead to systematic changes amounting to several percent in measured values.
In addition, there may be systematic errors in the calibration curves due to errors in the analysis values of the calibration samples.
Random errors are caused by changes, which cannot be checked, in the measuring equipment, environmental factors and the measurements process. It is well know that measurement results obtained by an observer with repeated measurements under the same conditions will differ slightly. The results are distributed around a mean value. Since statistical calculation procedures can provide information on random
Errors these are known as statistical errors.
Statistics are used as an important aid in assessing analysis results. Clear and unmistakable definitions help to avoid wrong interpretations and for this reason some concepts frequently used in analysis will now be explained.
Address: 19th Liangtong Road, Mashan Town, Binhu District, Wuxi City, Jiangsu Province, China
Copyright ? 2015 All Rights Reserved