Day 106 - 23 Mar 95 - Page 55


     
     1        one test per consignment, say they do 17 tests a day,
     2        something like 5,000 tests a year, if, as has been stated,
     3        they found E.coli 0157 twice since sampling was started
     4        about four years ago, what does that suggest to you about
     5        the possible presence of E.coli 0157 in terms of twice out
     6        of 5,000 or something?
     7        A.  None at all.  You cannot extrapolate at all one way or
     8        t'other.  All it says is out of those two consignments they
     9        found it. It means essentially they landed on it
    10        accidentally.  You could say that in those particular
    11        batches that it was high level, but an equal inference is
    12        that they blundered on it.  Unless you set up a
    13        mathematically designed sampling programme on a truly
    14        random basis you cannot extrapolate results.  Your results
    15        have no meaning for the purposes of extrapolation.  You
    16        cannot and must not draw conclusions from them.  There is
    17        no validity.  It is useless information.  All it tells you
    18        is on that day when you took that sample you found it.
    19        This is the trouble, because the nature of "random" is very
    20        poorly understood in general.  "Random" does not mean, oh,
    21        picking here, here and here.  It does not mean a person
    22        going in with the cutter and dunking it in where they feel
    23        like it.  That is not random.  Every person, whether they
    24        realise it or not, may be guided by their own inherent
    25        pattern; they make take it there, there, there.  That is
    26        not strictly random.
    27
    28        To set up a three-dimensional random sampling programme is
    29        incredibly complex, because you have to do in it 3-D.  What
    30        you would practically, there are several ways of doing it,
    31        but the way I would approach that is you would draw a grid
    32        over the box and then a vertical grid, so that you had a
    33        number of boxes.  You would allocate each of those a number
    34        and then you would run those through either a computer to
    35        give you random numbers or through random number tables so
    36        you had a box selected totally at random which you would
    37        then sample.
    38
    39   Q.   Are you suggesting that it is something that could be done
    40        once in a while?
    41        A.  Well, where -- I mean, in the real world, where
    42        obviously that type of sampling is expensive and
    43        disruptive, I tend when I am involved in this sort of work
    44        to put my money into sampling of very few batches very
    45        intensively, rather than sampling a large number of batches
    46        in a vague distribution.  I would tend to link it to
    47        setting up operations, but in terms of making claims which
    48        we have seen in Mr. Walker's, or satisfying expectations
    49        vis-a-vis McDonald's as to product being free from X-Y-Z,
    50        the sampling programme conducted does not enable the 
    51        supplier of that product to make any such assurance. 
    52 
    53   Q.   So if Mr. Walker said in his evidence that the meat was
    54        free from E.coli 0157, what is your conclusion on that?
    55        A.  If he said that, and I am not sure he did say that,
    56        I think he said that McDonald's expected it to be free from
    57        0157, all one can say of that is that McKey's have no way
    58        of knowing whether or not it is free from 0157 and are not
    59        carrying out any scientifically based programme which would
    60        enable them to give any such assurance.  In fact, the only

Prev Next Index