DKPOmiOKS AM) FXi'LANATioNs. xliii 



omitt'u. siijn, divided by their nun.ber) will be called the average 



will be denoted by ad. 



This tjiianti- . i-y tin.- sonaiv root of ,/ w,,uld !>.- -died the averagt 



tion of the mean, and will be denoted by AD. It is the average amount by 

 which any OIK- su.-h iiu-an would be found to deviate fr<>m the average of a 



means- all taken undrr like conditions. The term d> 

 measit rrini; to eithrr of these quantities. 



My, then, the n. may be summarized as follow - 



ilt is of a (iuoted amount 



is meant that tin- ;;ited or found to be of 



Miount. and that so far a- iy inspection, no 



: exist which a suit by an amount sensible a- 



compared with this.* 



It is essential to bear in mind in connection with all quantities that are the 



t no absolnt< numerical 



>r is possible; that any expression which is isually in- 



>n measure; that is, an approxii i^e value of the e; 



. accompanied by an assurance, expressed or 



i. that a Jl Me discoverable s.uirees of er: ess has 



. and that I been corrected f..r or f.mnd to ! 



the deviation measure. 



' he accuracy d->iivd in the result, it must be understo.xl that 

 the converse of this is meant; or, at least, that only the converse is possi- 

 ble of attainment. 



Rules for Significant Figures. The rules are so framed that, barriim mi- 

 the greate- -.putation suit of an\ 



ordinary computation (e.g. one involving a total of not in . linn 20 



component numbers, steps, or operations, where a rejection error may occur) 

 shall not be sensible compared with U Of data, n 



shall not sensibly atYrct the accuracy of the n-sult. They a: re, safe 



ssible cases. Hut in onler to be so they are 



more than suffiei' >nme classes >f coniparativ.'ly 1-011-: 



where -ranee of a computa; 



times as large as the expcM h \\-rk 



H place of figures may be \vln -n the ruli s are thus relaxed the 



possibl- uce slum: I iii mind, and special scrutiny applied to 



the various stages of tin- computation, spceial attention being directed to *|iian- 



bt'itfnninn with i n 

 Rejection Error. \Vhene\vr it br.-omcs necessary to throw off 



reason the last two figures are to be rejected in 



2437529L 

 making it 24375300. 



+ 9> 



' is always teased by i wh< 



rejected flgu > ,M- >i if that figure is less 



a more extended discussion of this and allied subjects see the an 



