editorial

2
Polymer Testing 8 (1989) 229-230 Editorial Interlaboratory comparisons, as for example reported by Veith in Polymer Testing recently (7(4) (1987) 239-267) have shown re- producibility of quite common tests to be much worse than had been hitherto realised. This is worrying not only from the point of view of pure testing, but also for the validity of specifications and their limits. Some limits are probably tighter than the test will apparently reproduce. What is also frightening to the tester or standards delegate is how to improve matters. It is fairly obvious to see that one approach could involve a vast number of interactive comparison exercises which, quite apart from bulk, would involve horrendous costs. Who would pay? Industry is notoriously bad at putting money into standards and if there is one subject worse it is testing research--nobody really wants testing, it is a necessary evil, so please don't add testing research! To get back to the point; the initial approach must be to consider all the possible causes of differences. These could be listed as: material; preparation; adequacy of standard; calibration of machine; and opera- tor error. It is perhaps useless to speculate on the relative contributions of these factors as doubtless they will vary from test to test and circumstance to circumstance. There is also the likelihood of interaction between them. However, it should be possible in any particular case to make a reasonable assessment of the most important and most likely effects. Material variation will always be with us, but samples can be single- sourced and distribution of the property accurately estimated within one laboratory, which should be the procedure for any interlaboratory test where the method is being investigated. The preparations, includ- ing variation in dimensions, can be included in the single-source procedure. Conditioning is a factor which may be judged very important for some materials and tests but unlikely to be a major cause of variation in other cases. The adequacy of the standard as written is probably the most difficult area to assess: if you knew the source of the variation, you would not be scrutinising the standard for what that silly other person could do wrong, nor in fact would this whole subject be raised! The questions are 229 Polymer Testing (8) (1989)----~ 1989 Elsevier Science Publishers Ltd, England. Printed in Northern Ireland

Upload: r-brown

Post on 21-Jun-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Polymer Testing 8 (1989) 229-230

Editorial

Interlaboratory comparisons, as for example reported by Veith in Polymer Testing recently (7(4) (1987) 239-267) have shown re- producibility of quite common tests to be much worse than had been hitherto realised. This is worrying not only from the point of view of pure testing, but also for the validity of specifications and their limits. Some limits are probably tighter than the test will apparently reproduce.

What is also frightening to the tester or standards delegate is how to improve matters. It is fairly obvious to see that one approach could involve a vast number of interactive comparison exercises which, quite apart from bulk, would involve horrendous costs. Who would pay? Industry is notoriously bad at putting money into standards and if there is one subject worse it is testing research--nobody really wants testing, it is a necessary evil, so please don't add testing research!

To get back to the point; the initial approach must be to consider all the possible causes of differences. These could be listed as: material; preparation; adequacy of standard; calibration of machine; and opera- tor error. It is perhaps useless to speculate on the relative contributions of these factors as doubtless they will vary from test to test and circumstance to circumstance. There is also the likelihood of interaction between them.

However, it should be possible in any particular case to make a reasonable assessment of the most important and most likely effects. Material variation will always be with us, but samples can be single- sourced and distribution of the property accurately estimated within one laboratory, which should be the procedure for any interlaboratory test where the method is being investigated. The preparations, includ- ing variation in dimensions, can be included in the single-source procedure. Conditioning is a factor which may be judged very important for some materials and tests but unlikely to be a major cause of variation in other cases.

The adequacy of the standard as written is probably the most difficult area to assess: if you knew the source of the variation, you would not be scrutinising the standard for what that silly other person could do wrong, nor in fact would this whole subject be raised! The questions are

229 Polymer Testing (8) (1989)----~ 1989 Elsevier Science Publishers Ltd, England. Printed in Northern Ireland

230 Editorial

whether all parameters have been specified, whether the limits are sufficiently tight and, most difficult, what variations in procedure may be unwittingly allowed.

Everybody always claims to have calibrated their machine but even where this is true we all know from bitter experience that mistakes can be made or important factors completely overlooked.

You cannot analyse operator error but all the possible systematic effects which come from error in calibration, misinterpretation of the procedure or erroneous practice will have been included in the areas considered above. Even the random effects of Monday morning are manifested in errors of calibration or procedure.

Perhaps the most interesting observation on an analysis of the possible causes of variability is that the best way of quantifying them may be to measure the effect of each one in a single laboratory. This is in effect saying that the best way to investigate interlaboratory differences is deliberately to do systematic trials in one laboratory. This route could be more efficient in isolating the main problems and, by spreading the work around several laboratories with one factor in- vestigated in each, it may considerably ease the financial difficulties.

R. Brown