How relevant are current standards of precision for on-farm milk meters and samplers?

 

'The standard is set ridiculously high,' is the conclusion of DPI Ellinbank's Tim Clarke after a detailed analysis of the various components of imprecision that are involved in herd testing.

 

Funded jointly by the Department of Primary Industries and the Dairy Herd Improvement Fund, the project examined the precision of various alternative testing regimes including some which can improve labour productivity on farms and reduce the hassle that farmers associate with "test day".

 

The International Committee for Animal Recording (ICAR) set standards of compliance for milk recording and sampling equipment.

 

At first sight the setting of such high standards seems commendable, but meeting such standards comes at a cost which is incurred by equipment manufacturers, service providers and in the end this is passed on farmers (who can't pass this cost on through higher milk prices). The model will be used to focus world-wide attention on the rational of current ICAR.

 

"We were asked to examine the relative contribution of milk recording equipment precision to overall precision of lactation production for individual cows," said Tim.

 

There are two main reasons for imprecision in herd-test results as a measure of 10-month lactation production.

 

First, with most herd-test systems production is measured only on one day per month. Thus the herd-test data are only a snapshot of the true production.

 

Second, on the days when production is measured, there is always some degree of imprecision in the measurement equipment.

 

This could be particularly important if equipment does not meet the ICAR standards. On the other hand it may not be very important given the much greater variation in day-to-day milk yields of cows together with once-a-month sampling.

 

Tim got together with Murray Hannah, the Biometrician at Ellinbank and summarized data on daily production variation in cows, and equipment precision. They developed a mathematical model that could use these to predict the precision of herd recording 10-month production estimates for various testing regimes when done with equipment of hypothetical specified precision.

 

The precision factors and other variables that went into the model were:

The model showed that the imprecision of the equipment at 2.5% coefficient of variation (as required by ICAR) was such a small contributor to overall precision that equipment precision could be relaxed two or three fold and only caused a small increase of overall imprecision. For example, the overall imprecision for 'Victorian standard herd testing' (once a month pm+am for the whole lactation) was estimated to be 5.07% coefficient of variation with ICAR approve equipment and this only increased to 5.27% if the equipment imprecision was doubled!

 

Many people are 'turned off' by statistics, but to understand this in practical terms an imprecision value 5.07% means the following. Let's say that the farmer's herd test results at the end of lactation show that his favourite cow had produced 150% of the herd average for that lactation. The farmer will know that the cow was probably 50% better than the rest, but because of measurement and sampling error, this cow's production could in fact be anywhere from 145% to 150% of the herd average. With equipment imprecision doubled (2.5-5.0%) there would only be an increase in the imprecision of only 0.2% units.

 

In discussing these issues with farmers and herd test service providers we have not found anyone who would be concern with such a minute increase in imprecision, particularly if it could make herd recording cheaper and more convenient.

 

A 'spin off' from the development of the model is that it can help us estimate herd testing precision for almost any imaginable combinations of frequency of milk volume recording and or milk sampling frequencies, coupled any milk meter or sampler precision. It should be a valuable tool for farmers (or their representatives) to use to determine what equipment precision/cost balance they may choose in the future.

 

For example, let us say that milk yield was measured daily using a low-cost in-place milk meter with even a three fold increase in imprecision (7.5%) and milk sampling was done monthly with an equally imprecise sampler (7.5%). This would result in 10-month production estimates for fat yield with an imprecision value as small as 4.83%.

 

This is a good example for farmers who are contemplating in-place milk recording. It shows how they can maintain recording precision while trading-off of equipment precision (and probably cost) against electronic recording milk yield every milking.

 

The model also shows that these farmers can also trade of the frequency of sampling (for their convenience) and still maintain overall recording precision.

 

The most commonly used testing regimes have been modeled and are available for examination. For those who may be interested in examining alternative testing regimes we can arrange to run your scenarios in the model. Contact Tim Clarke, [email protected] or phone 03 5624 2240.