106 



J. LlNDHAHD. 



We see here, that, with a single exception, for all the negative 

 deviations the barometer is 760 or below; for all positive deviations 

 it is 766 or more. Dividing the series into two groups on this basis, 

 we have, when the barometer is < 760, the average in mm. of mer- 

 cury = 300 4: 021, [J. = 0-75; when the barometer is above 760, 

 the average is ЗГ8 ± 018, ix = 067; the distinction is thus very clear. 



The amount of the carbonic acid per kilo, and per hour, 

 for the 12 experiments, is on the average 207 + 1'45, fi = 747 or 

 3-6 ^/o of the average. Including experiment No. 30 the average is 

 211-4- 2-8, n = 150 or 71 "/o of the average. Experiment 30 has 

 the value 257; its deviation from the average is thus -f- 46 or 306 /^. 

 The probability for a chance deviation of this order is 222:100000 

 or about 1 in 450. Calculating the highest probable limit of the 

 series according to the method already shown we have 242-5. It 

 seems justifiable, therefore, to exclude experiment No. 30 entirely. 

 If we also exclude No. 36, we have for average 208 + 146, pt = 

 720 or 35 **/o of the average. The deviations from the average in 

 the 12 experiments are distributed as follows. 



This again leads to the above-noted divisions and gives for 

 group I an average of 201 + 095, ц = 316; group II, average 212 

 j_ 1-56, n = 6- 14; In this last group, however, experiment No. 36 

 falls outside the calculated limit; excluding it we obtain for the 

 average 214 + 067, p. == 2-45; the standard deviation thus decreases 

 by more than half of its previous value, and the two groups thus 

 show themselves to be very distinctly separate. 



The following table includes 12 complete and one incomplete experi- 

 ment. This last was interrupted by the inspiration- valve of the valvular 



