|
CARDA's Computer and Biometric Services Unit, or CBSU, does not restrict itself to maintaining the Center's computing facilities. It must also develop a range of software tools for scientists who are performing a series of complex tasks. One of CBSU's operations is biometrics--the treatment of statistics to see what they are actually telling you (see Caravan No. 4). Another related, but separate, task is to assist the scientists with the decision support tools in the form of information management and statistical computation on diverse types of data collected from their experiments. It was the lack of sufficiently powerful and user-friendly software for this that led ICARDA to develop the Trials Management System--TMS. To understand why TMS is needed, and what its development involves, it is necessary to describe how and why trials throw up so much data. To the layman, a crop experiment must seem simple enough. One takes seeds of the variety one wishes to test, plants them in the appropriate environment and sees how they perform. But it is not that simple. The environments in the ICARDA region are diverse. A crop will face a huge variety of climatic conditions and pests and diseases from site to site--to say nothing of different soil types, fertilizer and pesticide use (or lack of them). Lines for possible release to farmers must be tested for all of these. So ICARDA collaborates with national scientists all over the West Asia and North Africa region and beyond to carry out multi-location, multi-year testing of advanced lines to identify improved germplasm. The heart of this cooperation is the international nursery system, which not only distributes ICARDA's improved germplasm but also functions as a cooperative testing vehicle. Candidate lines are evaluated at many key sites with stresses such as drought, heat, cold, salt, disease and insects. Data returned from cooperators provides valuable information on the performance and adaptation of test genotypes. Such efforts are an integral part of ICARDA's collaboration with the national programs. Every year, ICARDA's crop improvement program receives data from an average of 30-35 locations around the region for around 80 yield and stress nurseries with an average of 40 test lines. The information received includes performance data such as seed yield, total biomass and 100-seed weight, days to maturity and plant height. But it also includes data which could have a bearing on the test results--the fortnightly meteorological data, agronomic information such as amount of irrigation, types and quantity of pesticides, herbicides and fertilizer used, and damages due to pests and diseases, drought or cold. There is no suitable commercial software available that can be readily used to first manage the complex distribution system, and then process and produce the reports based on such a mass of information, correlating all the factors involved. Line X performed well at 14 sites out of 16 but was lousy at two. Why? The scientist's comparison must incorporate all relevant factors. Line Y incorporates good yield and stress-resistance characteristics from several previous lines, but was a mediocre performer. Why? What's missing? Attempt to answer these questions from piles of computer printouts, and you will never have time to do anything else. And you may still overlook the answers. A tool was needed not only for statistical analysis, but also for administration of the collaborative testing system. Now CBSU has developed a powerful user-friendly software tool called Trials Management System (TMS) to automate the various functions of the international nursery system.TMS manages and produces the reports from the relational data of test lines and its parentage, information on experimental design and the field plan, site-specific information including the meteorological information, agronomic management and the observed attributes using a Relational Data Management System (RDBMS). The statistical analyses on the stored data are performed using commercial SAS (Statistical Analysis System) software. TMS correlates all the information, which will be gathered by the scientist in the field (often using a palmtop computer). TMS is able to report on a given line, or a group of lines at a given location or across all locations and years; it can present this information graphically if required. TMS has other analytical tasks besides comparing lines under different conditions. For example, in a large number of field experiments there is bound to be the odd gap; a specific test line was not planted on one site, or the plants of a variety in a plot within a replication got damaged when a flock of sheep got loose in the field. Statistical analysis of such missing data requires special treatment (statistically expected mean with adjusted precision for that mean). This is difficult; a standard spreadsheet could "fill the gaps" by working out a simple average, but for the scientist this is not good enough. The standard margin of error from the other figures must be incorporated. Also, it would be very easy to analyze genotype by environment (G x E) interaction of the lines tested using data stored in TMS database. This is the extent to which
|
|