1 46 Standardisation o£ the Vitamins 



its cure. The length of time taken to effect a cure is therefore difficult to 

 estimate. Even the increase in weight is not very accurate, for a rat's weight 

 varies almost from hour to hour and may not be the same (or increase steadily) 

 at the same time on successive days. 



3. Adequacy of the Basal Diet.— ¥or a simple comparison between two sub- 

 stances (for example, the standard of reference and a food substance) with 

 respect to a particular vitamin, it is obvious that the diet of the animal used 

 should contain abundance of all substances known to be necessary for the 

 well-being of the animal, except the one vitamin which is to be assayed. 

 Otherwise, the food substance or even the standard preparation may supply 

 some factor which is helpful to the work of the vitamin under assay and the 

 restdts may thereby be vitiated. A slight deficiency of one of the vitamins does 

 in fact limit the steepness of the curve of response to another vitamin.' 



4. Statistical Examination of Results.— More and more workers are submit- 

 ting the results of their experiments to statistical examination to determine 

 how much confidence may be placed in them. The accuracy of an assay de- 

 pends on three factors: (1) the standard deviation of the individual observa- 

 tions, (2) the number of animals from which the average result is calculated, 

 and (3) the steepness of the curve relating response to dose of vitamin given. 

 It is evident that the less the individual results differ from each other, the 

 greater confidence one may have in the reliability of the average, and that the 

 larger the number of animals used, the more nearly will the average approach 

 the true value. It is only necessary to draw two curves of response, one much 

 steeper than the other, to discover that a particular standard deviation of a 

 result corresponds to a smaller deviation of dose in a steeper curve than in a 



less steep curve. The significance of a difference between two means is deter- 



fji j^i 



mined by a simple formula /= — ^ " and by reference to a table of t 



V^i' + ea^ 

 which indicates the probability of the difference being significant.* When ^=3 



and the number of animals used has been about twenty, it is generaly consid- 

 ered that the difference found is significant. Some workers of a more optimistic 

 temperament are satisfied with a value of 2 for t. It can only be suggested that 

 these should repeat their experiment on the same or similar lines. 



The selection of doses for an assay leaves something to choice. If any dose 

 produces a maximum or nearly maximum result, or a minimum or nearly 

 minimum result, it cannot be used in the calculation of potency. Some workers 

 like each experiment to be self-contained and to give one group of rats a dose 

 of the standard and several different groups of rats different doses of the sub- 

 stance under test. They then equate the dose of standard to the dose of test 

 substance which gave the result most nearly equal to that given by the dose 

 of standard. This way of calculating the result makes no use of the informa- 

 tion furnished by all the rats in the other groups which were given other 

 doses. The defect can be remedied by drawing a smooth line as nearly as 



