Well, i was challenged to show the accuracy of provisions within this British Standard and was told to check my sample specimens.

The finding from this exercise shows the flaws of the standard in making approximate guesses regardless of the approaches or alternative methods. The standard only provide proximity of sample range rather than precision.

Approach A with more cores will definitely reduced the in-situ strength when you get a range of high standard deviation in comparison to Approach B which is based on (guessed but) standardized standard deviation among the three specified range of samples.

MS 1242 which is the local version of the extended BS EN 13791 which comes with Table 5, which speaks about checking the validity of the lowest core result by using ratio which is part of a descriptive statistics. What make things worst at their disposal based on this local specification when making assumption of standard deviation of 6% by the mean of the sample population. At the same time, the bias of the reading should not be remedied based on the discretion of the specialist if the finding construed the factor reduction due to other factors. This mean, one have to eliminate their lowest core result but then it is okay not to; cognitive dissonance if not oxymoron!

Let us skip the issue about the badly authored local specification and focus on the British Standard. When samples are compiled into SPSS, the result should be generated using P-P Plot and Q-Q Plot for validity and reliability for checking. This will allow us to see the relationship between distribution of empirical vs theoretical based on normal distribution curve.

Similarly, a set of T-test using unequal variance should be conducted to check if there is significance between the sample readings and the target strength. This will justifies the significance as per Annex C which described the possibilities of the overall population's standard deviation after conducting the test.

The result generated based on the statistical analysis indicates that the concept of normal and asymmetrical distribution is not fully acceptable in this exercise, thus provide a much lower predicted value than the actual statistically analyzed value. When checked, the value is more than 99.99% which is more than 5σ or 6σ which is ridiculous.

The reason behind this ridiculous result is due to the distribution curve which was adopted, where there are kurtosis and skewness - led to standard error.

With that being said, data should not be erased or eliminated if it is proven to have significance to the overall test. It tells a tale and it provides reason to explore. I believe it is time for many to start questioning the validity of every specification or code of practice. It is time to probe rather than to utilize what are provided blindly. But then again, specifications are for those who are not bothered to understand, to complete their task timely to a level of acceptance, and never about fulfilling their passion in learning.