We have experienced that causation, especially in agricultural phenomena, is complex and that the method of analysis used in natural sciences is not satisfactory in all respects. Some directives are given to disentangle this complexity based on the following ideas. The first point is connected with the thought that also in agricultural research with its applied character the hypothesis expressed in a model and followed by testing should supply the main contribution to new knowledge. As we have experienced, this is frequently forgotten. The second point is the idea that testing can also be carried out with observational data from experiments without artificial change (non-manipulative experiments). The third point is the knowledge that the research worker can choose out of many models and functions. In this it is not necessary to confine the choice to functions with few factors and to models in which the ceteris-paribus principle must be assumed. A definite advice which attack and which models and functions should be chosen, can not be given. Each problem requires its own method of analysis, each research worker should follow his own way and chooses his own models.

The use of the normal regression model to interpret the relationships between soll factors and plant characteristics such as yield and mineral content is open to certain objections. One of the most important of these is connected with the imperIections and limitations of the regression model used. In such a model the assumption is made that the so called independent factors do not influence each other, in other words a change in one factor does not result in a change in another independent factor. In many cases however, this assumption is not valid; this is particularly the case with investigations into the mineral relationships of plants.The investigation of S l u i j s m a n s 10 into the relationship between the MgO and K~O contents of herbage on the one hand and soil fertility and other factors on the other provides orte of the many examples in which the use of the regression model is not correct. Data obtained from an uncontrolled experiment in which the soll factors were not changed experimentally ( F e r r a r i a) were fitted to a regression model, in which the MgO or the K20 content of the herbage were assumed to be the dependent variables, and the MgO and K20 content of the soil, the humus content, the proportion of weeds and the crude-protein content of the herbage the independent or causal factors. This model is illustrated in Figure 1 in which the assumed causal relationships are represented by arrows, directed towards the effect. The influence of the eorrelations between the independent factors is eliminated in the statistical analysis and the assumption is made in this regression model, --8 1 --

No abstract

In agriculture estimates are often used. In this respect an estimate means the result of determining a quantity in which method either a scale is used imprinted on the mind (mental estimate) or the quantity is derived indirectly by means of a number of secondary characteristics (correlative estimate). The use of estimates has its advantages especially in economizing time and money ; sometimes it is the only method of determining the characteristic (e.g. the quality).The usefulness of the estimates is determined by the degree in which they represent the real value. A method is given to determine the standard error of eye estimates of the yields of grassland and oats. The estimates given in relative values running from 1 to 100 must be converted into absolute values. The calibration and the investigation into the accuracy of the eye estimates belong to the field of line fitting in which the weighed values and the estimates both are subject to error. The total errors are splitted up in accidental errors and in systematic errors on account of the subject and on account of the object. It is possible to diminish only the first two errors by replicated observations.

The usefulness is discussed of Boer-Ferrari's vegetation-unit classification (S. & F. XX [939]) as an index of yield, use-value, moisture supply and possibly fertility. (Abstract retrieved from CAB Abstracts by CABI’s permission)