[inline]When not to use treatment: predictor variables that are not based on distinct managements, or that are distinguished by information already contained in the trait table (e.g. site, cultivar, date fields) should not be given distinct treatments. For example, a study that compares two different species, cultivars, or genotypes can be assigned the same control treatment; these categories will be distinguished by the species or cultivar field. Another example is when the observation is made at two sites: the site field will include this information.
A treatment name is used as a categorical (rather than continuous) variable: it should be easy to find the treatment in the paper based on the name in the database. The treatment name does not have to indicate the level of treatment used in a particular treatment - this information will be included in management table.
It is essential that a control group be identified with each study. If there is no experimental manipulation, there is only one treatment. In this case, the treatment should be named 'observational' and listed as control.
To determine the control when it is not explicitly stated, first determine if one of the treatments is most like a background condition or how a system would be in its non-experimental state. In the case of crops, this could be how a farmer would be most likely to treat a crop.
Example of data entered into the treatment form for a control (left) and treatment (right) |
David 2011-10-07