360 MR. S. M. JACOB ON 



from these values being well within the limits of the probable errors. The average 

 coefficients of variation for the Kharif and Rabi crops respectively are roughly 25 

 and 44. 



Turning to the coefficients of correlation the relative smallness of the kharif 

 coefficient as compared with that for the Rabi is a very important fact. The values 

 seem fairly consistent, except that for Pasrur they are in both cases far lower than for 

 the remaining Tahsils. This occurrence is repeated for the kharif harvest when 

 dealing with the unirrigated areas in the whole of each Tahsil ('*>.'■'§ 4 seq.), where 

 the coefficient of correlation for Pasrur is so small that it is exceeded by its 

 probable error. I am not aware of any exceptional circumstances in this Tahsil 

 which would account for this lowness, and it is not necessary to do more than 

 to point to the probable errors of the coefficients, and the great reduction in 

 the values of the coefficients produced by inaccurate measurements. This latter 

 question will be considered more fully later on, as it has a bearing on statistical treat- 

 ment where there is some uncertainty in the data. 



However, the coefficients of correlation for the Rabi harvest are decidedly larger 

 than those for the kharif, and at first sight»this result would appear to be contrary to 

 expectation, as it was well-known that rain falling at the close of the period April 

 to September has, as a rule, a beneficial effect on the succeeding Rabi, whereas the 

 influence of the rain which falls from October to March is generally disregarded as a 

 factor influencing the kharif crop. In excluding the rainfall at the end of the period 

 April to September in forming the variable with which to correlate the Rabi harvest, it 

 might have been anticipated that a serious diminution in the correlation would 

 result. As a matter of fact, as will be shown later, this actually is the case ; but the 

 correlation of the Rabi crop with the total rainfall of September to March is so high, 

 that even when reduced by excluding the September rainfall from computation it 

 still remains higher than the correlation of the kharif crop. 



This difference in the coefficients of correlation appears to be significant, but at 

 the present stage it is a somewhat difficult matter to explain it. Without attempt- 

 ing to give a complete explanation certain considerations will be adduced which will 

 show in what way the coefficient of correlation can be altered, some of which may 

 partially account for the above-noted difference. 



In the first place neither the crops nor the proportions of the principal crops are 

 the same for the two harvests. Rice, maize and jowar (Indian Millet) are the chief 

 crops of the autumn ; and wheat, barley, gram of the spring harvest; but whereas 

 wheat is, as a rule, more than half the total spring crop, jowar the largest of 

 the autumn crops rarely forms more than one quarter of the total harvest. 



Secondly, there is the important matter of the distribution of the rainfall. Clear- 

 ly if we correlate the amount of crop and the total rainfall, then whenever the dis- 

 tribution of the rainfall varies, the correlation will be diminished. 



This is an obvious deduction from the well-known fact that the same quantity of 

 rain will have a completely different effect on the harvest according as it its distribu- 

 tion is beneficial or adverse. 



