GRADING
Grading is a fairly important activity. The grading system used in Tanzania is modelled on the USA system, and on Study and Analysis by the Texas A & M University in Tanzania a few years ago. The tables are reproduced in the appendix of this report.
According to these Texas A&M Tables, for a given liveweight of animal, there is an approximately 10% variation in edible weight (muscle plus fat) for each change in grade; thus if a data recorder makes a mistake and misclassifies an animal by one grade, then there will be potentially a 10% error in that data point.
There are several ways in which grading errors can be minimised:
- try to ensure that only one person (the same person) grades all animals at a market over as long a period of time as possible
- organise a course in training for graders, and especially conduct a live blind test for all 3 or 6 graders (and for the course tutor) at a cattle market (e.g. Pugu) before the start of the course and again at the end.
- encourage the grader to use half-grades or even quarter grades if an animal appears to fall between 2 grades; also of course modify the computer program to accept and to process using these grade decimal fractions
- on the data sheets, have a field or column for the ID of the grader, and ensure that the computer system has provision for capturing that data.
- analyse historical, recent and current data to try to ascertain the grading performance of various graders, from the above information.
Note that in the original system, the grades were used (like the sex) as a classification only for which average unit price information was calculated; in the log-log system proposed in this report, both grade and sex are used actively with the Texas A&M tables to perform curve smoothing to increase accuracy and potentially reduce the sampling requirement.
Note that it would be very desirable to access the raw data recorded during the Texas A & M study, in order to properly determine the truth of the statement above that a one point grade error results in a 10% error in edible weight. I recommend that TLMP should approach the Texas A&M or adopt other channels to get
access to this data (the 10% estimate may actually be only a 5% in reality).
Grading errors can be probably up to +/- 0.5 grade, which means an possible error of up to +/- 5% on edible weight. The main problem may in fact be errors between markets, with data within markets being relatively tight.