STATISTICAL TREATMENTS 211 



tions. The assumption that the samples are random is a very important 

 one but one too often forgotten. 



Statistical significance means only that the conclusion drawn is prob- 

 ably correct. Biometry is nothing more than a formal procedure for 

 calculating probabilities. If the experimental data are good, that is, if 

 the error or variability is small relative to the effect being measured, there 

 is no need to analyze statistically. Statistical procedures are particularly 

 valuable in those experiments where the errors are large or the effects 

 are small. 



Analysis of results and drawing conclusions require common sense. 

 Blind acceptance of statistical significance can lead to ridiculous con- 

 clusions. I remember one experiment in which three rows of bean 

 plants were given identical experimental treatments, but one row pro- 

 duced more beans than the other two. The experimenter said, "I don't 

 know what it means, but it was significant at the 5 per cent level." 



Most biologists are not sufficiently expert to know which test to use in 

 every circumstance, and most cannot remember all the assumptions that 

 went into the development of the various analytical procedures. It is 

 a common practice to enlist the help of an expert statistician or bio- 

 metrician. The time to do so, however, is before the experiments are 

 performed. Most biometricians prefer to assist in the planning stages 

 rather than being called in at the last minute in a desperate attempt to 

 salvage something from a ruined experiment. 



SELECTED REFERENCES 



Fisher, Ronald A., Statistical Methods for Research Workers. Edin- 

 burgh: Oliver and Boyd, 1950. Sir Ronald Fisher's name is at the 

 head of every list of persons who have contributed most to biological 

 and agricultural statistics and to experimental design. This book is 

 a classic and still one of the best available for reference. 



Freund, John E., Paul E. Livermore, and Irwin Miller, Manual of 

 Experimental Statistics. Englewood Cliffs, N. J.: Prentice-Hall, 

 Inc., 1960. This little book will be convenient for those people who 

 understand statistics but cannot remember the details about equations 

 and such. 



Snedecor, George W., Statistical Methods Applied to Experiments 

 in Agricidture and Biology, 5th ed. Ames: The Iowa State College 

 Press, 1956. One of the standard biometry textbooks, with good rea- 

 son. Useful as a reference because it is complete, even if you must 

 sometimes do some reading to find the exact detail you need. 



