What do you consider a sample to mean during an EICR

hmm!  I think maybe some folk are looking at the wrong end of the stick I was trying to point with. Generally, and especially domestically, we do not know anything  about who has done what so the independence or correlation between looking in one place and knowing the quality of workmanship in another where we are not looking could be almost anywhere between zero and 100%, either way..

At which case the whole sample 20% of the circuits if you have more than 100 to look at has no more mathematical basis than a game of pin a tail on the donkey. By all means do it, but be clear it is not backed up by sampling theorems, the central limit theorem or anything else out of the A level applied maths and statistics course - like assuming that only low impedance faults occur, it is simply  a fudge to make the cost benefit balance look better, but one that is better than nothing and gives reasonable coverage in practice. In general I suggest that poor work is clustered, not randomly distributed, and tends to form a trail behind poor workers.

When did you last roll the dice when wiring a fitting to decide if this was the one where you were not going to bother to tighten the earth screw, or to wire it reverse polarity ? I'd be amazed if anyone did... But only by that sort of method is bad work truly randomly distributed. And only then do the stats work properly.

Mike

mapj1 said:

At which case the whole sample 20% of the circuits if you have more than 100 to look at has no more mathematical basis than a game of pin a tail on the donkey.

I never said "circuits", I said "population" - and provided an example of how coverage could be met. In fact, I'm agreeing with the statement that 20 % of circuits has no real meaning.

The approach has to be a coverage that provides confidence overall, using a combination of inspection and testing.

I think we are misunderstanding what sampling (in large installations) really means. My take is that say 20% should be examined (tested and inspected fully), and depending on the findings, and if they are problem free we can reasonably assume that the other 80% are generally satisfactory. However, and this is the important point, there can never be any guarantee that this is actually true. The question really is "does this lack of confidence matter?", and this is entirely down to the client. In some high risk situations it may matter seriously, in the average relitively low risk situation (using the accident statistics) it may well be decided that this an acceptable level of risk.

The problem we have in present times is that there is probably NO acceptable level of risk, and this is a cultural change over the last 50 years or so. This attitude has a serious downside, that risk and levels of risk are no longer understood due to lack of ability to assess properly. Risk is always present, it is part of life, but many now think that everything is risk free or must be made so, in my view to the detriment of life itself.

I notice that even very low risk jobs on domestic houses are being said to require full scaffolding, whereas a few years back a ladder was fine. The effect is that costs have gone through the roof, there are not enough scaffolders, and scaffolding is a relitively high risk occupation so all that has happened is that accidents are transferred to scaffolders! The industry now has quite a few cowboys, because of the above causes. Mad!

Sampling is guessing and assuming. Engineering works properly on absolutes not roulette. Just inspecting or testing part of an installation only provides part of an answer.

Four of my car tyres are at the correct pressure. The spare wheel is difficult to access, but as the majority of tyres are o.k. I will assume that the spare is as well.................until on a dark wet stormy night I find that it is flat. That is sampling.

Z.

Surely sampling is when widgets falling off a machine onto a conveyor belt are investigated, say in a one per 100 system, to confirm correct quality of  manufacture.

Definition...

Sampling is a process used in statistical analysis in which a predetermined number of observations are taken from a larger population. The methodology used to sample from a larger population depends on the type of analysis being performed, but it may include simple random sampling or systematic sampling.

Z.

Surely sampling is when widgets falling off a machine onto a conveyor belt are investigated, say in a one per 100 system, to confirm correct quality of  manufacture.

Definition...

Sampling is a process used in statistical analysis in which a predetermined number of observations are taken from a larger population. The methodology used to sample from a larger population depends on the type of analysis being performed, but it may include simple random sampling or systematic sampling.

Z.