| | There is growing need to have a basic understanding of statistics:
- for a clear description of experimental design and statistical analysis in a research proposal; and
- for a valid and better interpretation of research results.
Many researchers from within and outside Alberta Agriculture, Food and Rural Development (AAFRD) have annually submitted research proposals to Alberta Agricultural Research Institute (AARI) for research funding. The general observation during reviewing these proposals has been that while the objectives of a research proposal are usually clearly stated and the variables to be measured or the treatments to be compared are usually described in detail, what is often lacking is a clear and/or detailed description of the proposed experimental design and statistical analysis. Most often, the total elaboration is limited to just one sentence, such as "the data will be subjected to a regression analysis". It is no secret that AARI and many other granting agencies require proposals with a clear outline of experimental design and statistical analysis to ensure the quality of proposed research. Specifically, the two relevant questions have become part of the standard criteria set forth by AARI to evaluate the research proposals:
- Is the experimental design sound?
- Is the statistical technique to be used well explained and appropriate?
Therefore, in designing the research experiments and writing research proposals, researchers must know just what design, data and analysis are needed to support their proposed hypotheses.
With the recent advance in microcomputers and easy-to-use statistical softwares, more and more AAFRD researchers with varying degrees of statistical training are carrying out their own statistical analysis. There is a real danger of
- misapplication of statistical methods by the researchers with limited statistical backgrounds and
- invalid interpretation of outputs from theanalysis.
Furthermore, when mathematical difficulties arise from a research experiment, or when there is some doubt as to the best experimental design to be adopted, etc., the researcher needs advice from a statistician but he/she also needs to have some understanding of statistics in order to know whether or not the statistician is solving the right problem. The last thing that both statistician and researcher want to have during designing and/or analyzing research experiments is the so called "error of the third kind" (A.W. Kimball 1957, J. Am. Stat. Assoc. 57:153), that is, obtaining the right answer to the wrong problem!
The following discussion along with examples wherever applicable is organized into three sections:
This reflects the fact that statistics is involved in different stages of research: planning, experimentation and data analysis. The short discussion given at each section is aimed at helping AAFRD researchers to appropriately use statistical concepts in designing and analyzing their experiments, thereby enhancing the quality of their proposed research. The in-depth discussion can be found in references. |
|