Presentations at Fertcare NUE workshops

Gaining confidence in the information we provide

Two presentations were made at the Fertcare Nitrogen Use Efficiency workshops across the eastern states. Both relate to gaining confidence - firstly in the information we seek to share, and secondly with the process from soil test to nutrient recommendations.

PDF versions can be downloaded below.

When checking on the claims of a product, the first and most important thing is the evidence the supplier has about the crop response. This evidence should be done in a scientifically credible way using methods that is explainable and reproducible.

Appropriate controls – every fertilizer experiment should have a nil treatment (no added fertilizer) and a standard practice. Without these checks, there is no indication if the new product actually did anything, or if it was better than the standard treatment. Comparisons should be done at least on a nutrient to nutrient basis where similar amounts of the nutrient are applied so the comparative efficacy is clear.

Replicated – are the trials replicated, which means are the treatments at a particular site repeated so that the information collected can be statistically compared. Replication is the basis of establishing natural variation in an experiment, and without that, the effects of the treatments cannot be distinguished from luck.

Randomization – the treatments should be randomized in such a way that one is not necessarily in the same place in each replication. Often treatments will be blocked together so that paddock trends can be accounted for in the analysis.

Repeated – one trial in one year at one site does not give proof of a response. Has the trial been done on relevant soil types, in appropriate regions and on the same test crop.

Compared statistically – a replicated trial will have a mean (or average) and a measure of error for that mean. The error term gives a range of “normal” values for the mean so that the ranges of different treatments can be compared. Means are significantly different when these ranges do not overlap at a particular probability. If they do overlap, even though the numbers are different, there would be no “significant” difference between the treatments.

Scientists start with the premise that there is no difference between the treatments, and design experiments to test this. Endorsements and product testimonials are no substitute for good experimental design and robust statistical analyses.