Setting
A "Spec"
Home
Setting, checking and adjusting a Specification (Spec) are, in reality, decision making processes that can be very unnerving, but can also be very profitable if done well. The steps involved in setting, checking and adjusting a Spec are quite simple to carry out. What makes the process unnerving is the decision that has to be made as to the level of the Spec, and the risks involved for setting too broad, or too narrow a Spec. A basic understanding of statistics can take the uneasiness out of the decision making process.
The scope of this text is to illustrate how to set a Specification using the information supplied on the Caltest, LLC Certification Certificate. Checking and adjusting your Spec can be accomplished with the same illustration but will not be discussed in great detail here.
Some of the terms you may have heard of in other discussions of setting specifications may include some of the following: Precision, Accuracy, Standard Deviation, Degrees of Freedom, Sigma, Mean, Control Limits, Residuals, Population, Confidence Interval, etc. Only the term Precision will be used in this presentation.
Example:
Assume a nutritionist has determined that he needs a 50% protein soybean meal, measured on an "as received", or "moisture" basis to formulate his diets.
Assume also, that you are using an NIR as your analytical method to rapidly test your soybean meal for protein content, and that your analytical method has been certified under the Caltest Certification Program.
What Spec do you need to set in order to assure your nutritionist that he gets the 50% protein soybean meal from your supplier?
If your soybean meal calibration was developed on a dry matter basis, all computations to determine the specification should be performed on numbers that have been converted to dry matter basis.
Assume that your NIR analysis results for moisture in soybean meal have been running at around 11.5%. The Dry Matter (DM) is then: (100 - 11.5 = 88.5%).
Refer to the Caltest Certification Certificate.
The NIR Precision for our example, using the Caltest Certification Certificate, has been found to be 0.26% for moisture and (DM). This means that if you want to be 68.3% certain that the moisture, or (DM) result you will report for a given soybean meal is correct, you must report the result as a range, or 11.24% - 11.76% for moisture, and 88.24% - 88.76% for (DM). You will have a 50% chance of being right if you report your moisture or (DM) result as 11.5% and 88.5% respectively.
You will have a 95.4% certainty that the result you are reporting is right, if you report a range that is twice as wide: (0.26% X 2 = 0.52%) or 10.98% - 12.02% moisture, and 87.98% - 89.02% dry matter. A 99.8% certainty requires a range that is three times as wide.
Your nutritionist requested soybean meal to be at a 50% protein, moisture basis level. Convert the moisture basis protein to dry matter basis protein:
Protein (Dry Matter Basis) = Protein (Moisture Basis) ÷ DM X100
Protein (Dry Matter Basis) = ((50/88.5) X100) = 56.5%
According to the Certification Certificate, the NIR has a certified precision (error) for protein analysis of 0.48%. To be 95.4% certain that the result reported is right, we should compute the following range for protein:
6.5 - (0.48 X2)] - [56.5 + (0.48 X2)]
or [56.5 - 0.96] - [56.5 + 0.96]
and 55.54% - 57.46%
To set your specification on an "As Received", or "Moisture Basis" one must now consider the error of the moisture results calculated earlier.
Minimum Protein Level = [55.54 X (87.98/100)] = 48.86%
Maximum Protein Level = [57.46 X (89.02/100)] = 51.15%
To assure that your nutritionist gets a 50% protein soybean meal, on an "As Received Basis", 95.4% of the time, your supplier can deliver soybean meal that will test out on the NIR, at a range of:
48.86% - 51.15%
One can reject loads of soybean meal if they fall under the 48.86% protein level after considering the possibility that your supplier can reject your claim 4.6% of the time.
You may need to consider special storage conditions if you notice a run of soybean meal consistently testing out greater than 51.15% protein, on an "As Received Basis".
Precision of the analytical method, in this case the NIR, was used in setting the specification. The precision of the referee laboratory could instead have been used.
Accuracy is used to examine the error of the "new" analytical method relative to the original analytical method. Accuracy comes into play when a specification needs to be checked, and will be discussed in a subsequent presentation.
Copyright Caltest, LLC 1999 All right reserved.