Should a farmer use a statistician to help him get at the truth in a farm trial? I’d say yes for two reasons. First to design your trial so as to get a result that can be relied upon; and second to interpret the results accurately and to give you unbiased advice on the trial of a product that you are being pressured to buy.
Very probably 80% of on-farm trials are worthless because, in not being set up correctly (that is statistically), the results invariably lie within the laws of chance and are therefore meaningless. Even more damaging, they can be misleading, causing producers to take wrong decisions. I’ve seen this many times in my advisory work.
Any trial should have a “will make me change” target, usually based on an economic return. A statistician is needed to set up the trial circumstances where this trial target can be reached – if it exists. Then the outcome is simple, you either can change if it’s positive, or don’t if it’s not.
If his advice involves too many pigs or replicates (which require the number of groups of pigs needed to get a valid comparison) plus the accommodation needed for them and the extra labour, then you have two options – ether reduce your “make me change” threshold to make it feasible for you, or abandon the project.
Hard words, but please stop fooling yourselves as so many do. Statistics are immutable – seek statistical advice before you start, and again at the conclusion to tell you what the figures say and don’t say (which is equally important) about the result.
So, how much does a statistician cost? The result of a farm trial is statistically positive, negative or inconclusive. Below I note the results of 17 farm feed trials carried out by my clients between 1998 and 2013. Eight cases were positive and nine negative or inconclusive (no change made/not worth the bother).
Postive result
For the positive results, the cost of statistician in relation to 12 months’ increased the gross margin (in five cases) or increased income (in three cases).
Farms: 1, 36%; 2, 1.6%; 3, 10%; 4, 0.7%; 5, 4.1%; 6, 3.0%; 7, 5.2%; 8, 4.3%. The average of all farms was 8.14%.
If the trial result is statistically favourable, then the cost of the statistician to achieve the result, on average from the above sample, eroded less than 10% of the gross margin benefit gained. (In the above group of trials this was an REO of 8:1 – and that’s excellent, a good growth enhancer can yield an REO of 5 to 6:1.
Negative result
We can repeat this exercise for a negative result. We can look at the cost of a statistician in relation to the current annual margin over food costs/pig, that is nominal money “wasted” because the trial didn’t meet the farms’ targets, so no change was made.
Farms: 9, 1.33%; 10, 3.81%; 11, 2.56%; 12, 1.89%; 13, 2.5%, 14, 1.7%; 15, 2.1%; 16, 1.9%; 17, 3.1%. The average of all farms was 2.32%.
Even if the trials were negative, only a little more than 2% of the statistician`s cost was “wasted” in terms of reduced margin over food costs. In reality his time is not wasted, as a statistically negative result warns you that it is not worth changing – assessed by your own standards of value for money.
I suggest that using a statistician for farm trial work is likely to be a good investment financially, as well as having peace of mind from not wasting your time. You can obtain results in which you can be confident and thus make better decisions.
