In ANSI N13.15-1985 (PERSONNEL DOSIMETRY), the criteria for Accuracy(Bias) and Precision(standard deviation) as follow:
for photon from .3 to 10 MeV: (Bias
Regarding accuracy, the best method is to place the dosimeter(s) in a know radiation field, for instance nearby a radioactive source of know activity, and to compare the TLDs reading with the known dose.
Regarding precision, the method is easier: place several TLDs in the same spot and evaluate the standard deviation of the sample.
Hello Majid,
The concepts of bias and precision are usually applied to systematic and random errors, respectively, associated with measurements. Accuracy, which measures the total uncertainty (systematic + random) between the true and apparent values of a measurement, is the sum of the bias plus precision. But while precision can be measured, accuracy cannot.
Precision is a way of quantifying repeatability of a measurement, i.e., if you are using TLD's, for example, you can estimate the precision of the measurement by exposing N dosimeters of the same type (same model number and same lot number and stored under the same conditions including time), one at a time, to the same radiation field and calculating the sample standard deviation of the dosimeter responses, but there is a rub or two in doing this procedure.
First, you need to convert the estimated sample standard deviation to an estimated population standard deviation, i.e., the sample standard deviation must be multiplied by a coefficient which takes into account the size of the sample, N, and the underlying distribution of the population of all dosimeter responses, not just the N dosimeters you employed. Note, it is very unlikely that the population distribution is normal (Gaussian), although the distribution of sample means may approach normality. The underlying distribution of the population of all dosimeter responses of the same type cannot be normal because the responses of the TLD's cannot be negative; a Gaussian distribution extends from negative infinity to positive infinity. A better fit for the underlying distribution of the population of all dosimeter responses of the same type would probably be a log-normal distribution.
Second, the precision you determine depends on a number of factors: 1) the intrinsic precision of the TLD's itself, i.e., the amount of active material in the individual TLD due to how it is manufactured; 2) the repeatability of the fixturing used to hold the TLD in place in the radiation field; and 3) the temporal variation in the radiation field itself.
Note, while precision quantifies repeatability (the spread in measured values from a single experimental setup and observer), it does not necessarily quantify reproducibility (the spread in measured values from different experimental setups and/or different observers), i.e., if a different researcher measures the precision of the same type of TLD's in a different laboratory, they will almost certainly find a different value of precision.
Accuracy, which is the sum of the bias plus the estimated population standard deviation, cannot be measured because the bias cannot be measured independently. Consider the following scenario. You place a TLD in a known radiation field and measure its response. The problem with the aforementioned procedure is the word 'known'. How was the radiation field known? You say it was known via a calibrated instrument, but how was the bias of that calibrated instrument independently determined? What happens is that you get into an infinite regress. Most people would say that you break the infinite regress by determining the radiation field in a national laboratory using a primary standard instrument such as a calorimeter or an open air ionization chamber, i.e., instruments whose response can be determined by measuring the mass, length, time, etc. of the detector. A TLD is a not primary standard, i.e., it is not an absolute instrument, it is a relative instrument. But even the bias of a primary standard instrument is only calculated. It should be noted, that national laboratories only use a very limited number of radiation sources. The radiation source (type and strength) you intend to measure with your TLD's may not correspond to the radiation sources used in the national laboratories. Hence, the instrument used to produce your 'known' radiation field might have its calibration traced back to a different radiation source than one being measured by your dosimeter.
I realize this answer is too terse, but I don't want to bore you anymore than I probably already have.
Regards,
Tom Cuff
- Badges
- Science topic
- Similar topics
- Econ
- Innovation Economics
- Innovation