VIII .
" On the Application of the Calculus of Probabilities to the results of measures of the Position and Distance of Double Stars .
" By THE LORD WROTTESLEY , V.P.R.S. , &amp;c. Received May 27 , 1859 .
In a communication addressed to the Royal Society " 'OnX the results of Periodical Observations of the Positions and Distances of certain Double Stars , " printed in the Philosophical Transactions for 1851 , I took occasion to remark that the differences between mean results obtainied on different evenings were greater in proportion than those of the separate or partial measures obtained on the same evening , which arise from chance errors of observation , and that this circumstance rendered the application of the Formulhe of the Calculus of Probabilities to the reduction of the observations embarrassing and difficult .
In other words , the differences between the mean positions and distances obtained on different nights were greater than would have been anticipated by one who had merely computed the probable error of a single measure in the usual manner from the data furnished by the sums of the squares of the partial differences from the mean .
The observations made since 1851 fully confirm the anomaly in question .
It is probable , therefore , that there is some cause which modifies sensibly and in some unknown manner the results obtained .
It may be temperature acting on the mierometer screw ; it may be the state of the atmosphere or the method of making the observation ; but whatever it be , the observations show conclusively that such causes are sometimes in operation .
For the purpose of obtaining some numerical expression , however imperfect , of the effect produced , I adopted the following method : I took the difference between two mean results of position obtained on two differenit nights , where not more than about two months intervened between the observations ; and I ascertained also the mean of the probable errors of such positions as computed in the ordinary method .
In order that each star might be subjected to exactly the same treatment , I selected always the observations of the first two nights on which it was observed , except when the two consecutive neans were obtained at too long an interval apart .
Now as the number of partial measures of angle obtained on each separate night very often did n-iot execed six , these probable errors are certainly not theoretically correct , or to be depended upon absolutely as a test of the accuracy of the observation ; but it may perhaps be assumed that any errors arising from this cause will not materially affect the mean of a very great number of results .
It appeared then that the mean of 218 differences taken at hazard from among such as were most accessible , and from observations made by different observers , was 374 67 , and that the corresponding mean of the means of all the probable errors was 13 ' 29 ; that is to say , the latter is 35 27 per cent. only of the value of the former .
As some proof that the cause , whatever it may be , is not very variable in its operation , I may add that the first 110 differences , which were all obtained before the end of 1854 , give these numbers , 37'*79 , 1 1"86 , and 31 37 per cent. respectively .
Again , the last 108 differenices , which were all derived from the observations of one observer only , give 37t 56 , 14t"75 , and 39-27 per cent. Of these 108 last differences , the first 50 , taken from the middle epoch of all the observations , give 40 ' 06 , 14'60 , and 36A44 per cent. , and the last 58 of the 108 give 35'O40 , 14'88 , and 42-03 per cent. There is , however , a circumstance which must be taken into account in making a comparison between the first 110 differences and mean probable errors , and the last 108 .
During the course of the observations from which the former were derived , it was the practice to take always 10 measures of each star on each night when possible ; during the observations from which the latter were derived , 6 measures only were taken .
This would tend to make the differenices less in the former case ; andlwith respect to its effect on the probable errors , if we put F for the error of a single measure in each case , the probable errors in the former case should be less by a quantity-0 00768 x F ; for if we put C for the constalnt and P for the probable error , then we have P= x F , where 6 measures are obtained ; and P= / 5 x F , t/ 4 4/ 45 where 10 are taken ; and the difference between these values =-0768 F. The facts above disclosed create the difficulties in applying the Calculus of Probabilities which have been before referred to .
In the first place , the partial measures obtained on each separate night are generally too few in number to eliminate the effects of onesided chance errors .
In the second place , it seems probable that some cause remains in action during a whole night , modifying the result , whose origin and law remain to be discovered , but which seems tolerably constant in its operation .
The observations of my Catalogue of double stars are drawing to a close , and it became extremely desirable that if there were any fault in the reductions or method of computing hitherto employed it should be speedily remedied , and the necessary corrections made ; I therefore applied to the Astronomer Royal , stating the embarrassments arising from the above-mentioned causes , and requesting his opinion as to the best mode of proceeding .
The Astronomer Royal exhibited on this , as on all other occasions , where his aid has been solicited , the greatest readiness to give me the benefit of his extensive knowledge of all that appertains to Astronomical science .
Mr. Airy observed that if there were a constant cause of error on any night , no multiplication of observations on that night would tend to remove it , and in that case he knew of no mode of proceeding which would quite meet the difficulty but the adoption of the following formulae .
Assuming that all the observations are equally good , or can be made so by grouping discordant measures , letf be the probable error of a single observation , and e the probable value of the error of each night .
Let S1 S21 S3 , &amp;c. represent the sums of the squares of the errors obtained in the usual manner from the observations on the first , second , third , &amp;c. nights respectively , and put n1 , n2 , &amp;c. for the number of observations obtained on each of those nights ; then the observations of the first night give , ( n-lI)f2=4549 xS , ; those of the second , ( n2 -l)f2= 4549 X SO , and so on : and from the sum of all these equationsf may be accurately determined .
Then to find e , compare the mean result obtained from the observations on all the niights , i. e. the mean of all the means , with the separate means for separate evenings ; then putting S for the sum of the squares of the errors founid by such comparison , and m for the whole number of nights , we have ( mn-1 ) e2= *4549 x S , which gives the value of e. If A be taken to represent any convenient constarnt , the combining weight for each of the first night 's observations will be A +f2 for each of the second night 's A , and so on .
n e2 +f2~ Let P stand for the probable error of the final result , then V(nie2 + n2e2 +f + 13e2+f2+ &amp;c. ) The probable error of the mean of the first night 's observations -V/ ( e2 +-C ) , of the second= V(e2 +.f ) , &amp;c. Mr. Airy , however , while remarking that the mode of proceeding above described is the only one which really meets the difficulties of the case , admits at the same time that it would not be expedient to use so elaborate a process in dealing with observations like those in question , in which the ordinary errors of observation are large in amount , and in which such extreme accuracy in the results is not obtainable as in some other cases to which the principles of the , Calculus are applicable .
He suggests therefore that all the observations of all the several nights should be combined together for the purpose of obtaining the probable error and weight of the final result ; and this may be dolle in two different ways:-First , by treating all the single measures of all the nights , as if theyl had been made on one and the same night , and obtaining the finial result and its probable error and weight accordingly in the usual manner : Secondly , by treating each group or set of 6 or 10 as a single observation .
The only other mnethod of proceeding is that above described as the correct one , but which has not been adopted , as being too cumbrous for the occasioni .
This will be designated as the Third Method .
For the purpose of ascertaining the result of employing each of these three methods , I requested myy assistant , Mr. Morton , to observe three stars , selected from among those which preseuit only average difficulties , a very great number of times , so that the measures should be sufficiently numerous to eliminate all onie-sided errors .
The observations of Position only have been used ; but these have been dealt with in the three different methods above described , that is , the final result and its probable error and weight have been obtained by each of the three modes .
The results are here subjoined , and the errors and their squares are given in full as to two of the stars , together with the whole computation ; and it is to be hoped that this may not only prove interesting to observers of double stars , but may throw some light on the curious mathematical question iiivolved in the inquiry which is the subject of the above remarks .
Among the stars selected as above mentioned for the trial of the three methods , was 2 Comae Berenices , or 1 1596 .
Now this star had been very frequently observed during the six years from 1843 to 1848 , at the time of the Parallax investigation , to which reference has been already made .
The comparison then made between the mean of all the measures of position obtained and the value of the angle of position given by Struve , gave reason to believe either that the angle had not altered during a period of sixteen years , or at least that it had altered very little .
The observations of 1859 fully confirmed this opinion .
Rejecting from the observations of 1843-8 those made on two iiights , when less than 6 measures were obtained , the result of 1859 differs only 8 ' from that of 1843-8 .
I was thus enabled , for the purposes of this inquiry , to treat these observation-s of 1843-8 , 156 in number , as if they had all been made within ali interval of time not greater than about two months .
Now these observations had beern made by three different observers , and while the results of separate nights were very discordant , the probable errors derived from the partial values of nights , the results of which differed greatly from the general mean , were as remarkably small ; on the other hand , the observations of the same star in 1859 , 215 in number , were by one observer only , and the results of different nights agree very closely .
The applications of the three methods to the early and late observations of this star therefore illustrate very strikingly the effect produced by discordancy in the valtues obtainied on different nights , when the pectiliarities of the object observed are elimilated .
Thus the good observations of 1859 give the e2 equal to 89 ' only , while in those of 1843-8 the es attains the great value of 8731t .
The values off ' given by the observations of the three stars accord very well , considering the different circumstances under which they were obtained .
It will be seen also that little effect is produced on the mean result by using these different methods of reduction .
In the account of the American Coast Survey of 1856 , and at pages 307-8 , will be found a formula by which the probable error is deduced from the differences from the mean alone , the probable error or P=0-845347 , , where e represents the error of a smgle observation .
I have tested this in the ease of three stars in which n was equal to 6 , 10 , and 156 , respectively , and the probable error deduced was a little greater in the first two cases , and a very little smaller in the last .
Computation of P by Method 1 for 2 1596 , 2 Com. Ber .
Epoch .
Sum of No. of Errors .
2 .
Errors .
A2 .
Errors .
O. Errors .
e6. .
Measures .
Meas..______ 36 I296 8z 6724 32 I024 53 2809 o/ 75 6z5 26 676 40 x6oo 137 18769 1859'238 22 29 I0 39 152I 115 I3225 21 44I 22 484 '24I 9 53 6 42 I764 59 3481 48 2304 i8 324 '244 21 43 I0 46 2II6 15 225 63 3969 19 36I '244 15 44 10 8 64 19 36I 23 529 20 400 '244 17 25 10 42 1764 12 144 6 36 24 '244 20 10 10 9 8I 29 84I 29 84I 65 4225 'z6o 13 31 10 76 5776 I67 27889 6I 3721 58 3360 *z6o I3 42 Io 42 1764 34 II56 83 6889 39 z6o 14 27 I0 58 3364 33 I089 14 I96 20 400 'z6o z6 I I0 23 529 53 2809 33 I089 9 8I 'z6o 20 48 10 54 29I6 68 4624 I4 I96 33 I089 'z63 13 z8 9 35 1225 94 8836 31 96I 86 7396 ez88 16 45 10 12 144 68 4624 7I 504I 79 624I 'z88 22 24 I0 5 25 25 625 88 7744 33 1089 '290 15 15 I0 151 zz8o0 35 I225 102 10404 100 10000 '293 II 40 I0 89 792I 39 69 476i 104 Io8i6 '293 i8 4 I0 6I 372I I3 I69 63 3969 35 I225 293 21 49 10 32 I024 13 16g 2I 44I 6o 3600o '296 I2 I 10 47 2209 78 6084 40 i6oo I0 100 '296 i6 z6 I0 34 1I56 13 I69 29 841 8 64 '296 24 30 I0 47 2209 39 152I 5 25 6i 3721 '296 21 35 I0 67 4489 I26 15876 I7 289 57 3249 zz_ _ 57 3._ __ 84 7056 78 6084 I0 100 17 289 22)I'547 389 50 ?
.zI5 5 25 58 3364 32 1024 58 3364 _ 49 I9 36I 79 624i I02 10404 7 49 I859.270 + 30 039 64 4096 48 2304 I7 289 49 2401 1G0 I0000 48 2304 20 400 3I 49 83 6889 37 1369 5 25 95 9025 Zero -450 I9 41 I681 35 I225 134 17956 23 529 -_____ II7 13689 94 8836 II 74 5476 4I8 30 5 25 39 152I 14 I96 134 17956 238 ? ?
30 ' 35 I225 34 1156 2I 44I 91 8z8i -Mean 00 25 625 54 29I6 82 6724 Position .
68 4624 85 7225 35 I225 4I i68i 56 3136 II3 12769 27 729 6I 372z 14 J96 26 676 86 7396 24 576 6i 3721 48 2304 54 29I6 51 260I I09 ii88i I05 l1025 65 4225 27 729 128 I6384 20 400 45 2025 127 I6129 32 I024 89 792z 52 2704 8 64 I8i 32761 54 2916 3I 961 92 8464 II I2I 14641 28 784 56 3136 I2 144 44 1936 85 7225 i6 256 77 5929 131 17I6I 34 II56 I09 Ii88i i6 256 31 961 98 9604 204 4i616 94 8836 87 7569 17 289 149 2220I 10 100 73 5329 00 103 o609o 149 22201 74 5476 94 8836 145 21025 76 5776 58 3364 34 1156 68 4624 69 476I 47 2209 27689i 92 8464 58 3364 25 625 152076 23 529 93 8649 23 529 266740 50 2500 35 Iz25 30 9o0 241952 24195'| |266740 15 076 S 62= 937659 Methlod 1 ( continued ) .
Log. of 937659=_59720449 Log. of 2 15 '=23324385 Constant =63995o98o Log. of 214 =2=3304138 P29'271 = o9671429 4 6628523 P3 '045 =o048357I5 Log. of 454936 = 16579503 W= IQ79 4 6628523 Constanit = 69950980 Method 2 .
Mean Position=z38 ' 30 ' .
I , ~~~~~~~~~~og .
of 12763 =4 I0595 Errors .
62 .
Errors .
62 .
__ ____ -Zn=22 , Constant=4 99331 26 676 89 361 x oz6= of p2 10 T00 8 64 I096 f~ 21 44I 25 625 o-54963=XOf P 15 225 17 2.89 P2=2.6 4 x6 39 152I p= 56 12 144 11 I1 " 4 28 784 22 z 484 W= '0796 27 729 37 I369 22 484 IO IOO 47 2209 38 '444 x6 256 zx 441 6064 6699 6064 2 62 12763 Method 3 .
15o63=Si 9f2=CSI 7583=S2 5f2 CS2 2zI8zo6=S3 39 f2-CS3 24i62i=S4 49f2= CS4 2.z04 =S5 8 f2-CS5 48 249 = S6 I,9 f2-CS6 Gf2=CS74 345 II= S7 9f -CS 29f2=CS ' 245830=Sg 39 f2CS , 92.0274 z06of2-C X ( SI+S2+ &amp; C. ) L0og .
of 920274 5,9639I72 Log. of C=4549 &amp;c. .=-i6579503 5-62I8675 Log. of zo6 z23I38672 f 2032'4 -3'3080003 Method 3 ( continuted ) .
bi=32 I5 bl-'=26 &amp; ( bl-b)2=676 b2-31 39 b2-b= Io ... ... ... ... o00 13=33 53 b3-b= 4 ... ... ... . .
I6 b4-31 46 b4-b= 3 ... ... ... . .
9 b5=31 30 b15--I9 ... ... ... ... 36I b6-31 57 b6-b= 8 ... ... ... ... 64 )7=3I 32 b7-= 1I7 ... ... ... ... 289 bs =3I 43 bs-b= 6 .
36 b)= 3I 52 b-b= 3 * ... * .
9 b-=3I 4.9 I 156o See Method 1 .
Log. of 2 560 3 193 1 246 Constant Log. = r65795o3 v-851 0749 Log. of 8 =m-I 0-9030900 e2=88`7 I I9479849 'V'{ ( I0xsX7 ?
+032 ) +(61 6 +2032)+(40 X 88+203 ) + ( X 887+2032)+ ( 9 X 88'7+2032 ) ?
( 20 X 88'7+2032 ) { I0 + 30 +| 40 +io X88 ?
7+2032 ) ( 30X887+2032 ) +40x88 7+2032 } ioX88'7887 50X88'7= 4435 + 2032 X I0= I +2032 A 50=I'69897 2919 &amp; X3 46523 6467 &amp; X = 3-81070 ISt w 00oo3426 -35347 7 4th W= *007732 =3-88827 6X887 = 532 2 9X887= 798-3 + 2032 ' A 6=0o77815 +2032 ' A 9=0-95424 2564R2 &amp; X -3340895 2830-3 &amp; X =3 45I84 znd W =0072340 =336920 5th W= *003I80 -350240 4oX88 7= 3548 20X88'7 1774 +2032 X 40= I60206 +2032 X 20=-30103 558o &amp; X 3-74663 3806 &amp; = 3'58047 3rd W= '007169 3-85543 6th W '005255 3-72056 Method 3 ( continued ) .
7th W o00342,6 3oX88'7= 266T _ +2032 X 30 ?
-147712 gth W '007i69 4693 &amp; X -3 67145 8th W= oo6393 =3 80567 ISt ='0034z6 X of x -0 znd= oo2340 X of 046og=z66361 3rd =0007I694th =007732 vt33639-X Of P2 5th oo003 i8o o668zo==X of P 6th '005255 7th = *00342z6 P-2 , ]1 70 8th= oo6393 P 4'`658 gth =oo70I69 W= '046I W ='o460go 310+ ai=75 Wi 34 75 39 53 46 30 57 a2= 39 W2= 23 34 23 72 77 32 53 a , = 53 W3= 72 a4=46 W4= 77 300 II7 106 322 60 I71 ai =30 W65 32 225 78 37I 322 go 285 a6= 57 W6T.53-a7=32 W7 34 2550 897 3816 3542 960 3oz2 as=43 Ws= 64 a 52 W9= 72 32 43 52 46I 34 64 72 IZ8 I72 104 96 258 364 1o88 2752 3744 2550 897 38i6 3542 96o 302I x088 2752 0 3744 Mean result:= 31 49 Zero 450 19 461)22370(48-5 -I844 418 30 Mean Angle -38 ? ?
30 ' 3930 3688 242O 2305 2 3049 , a Cassiopeae .
Computation of P. Method 1 .
Epoch .
Sum .
No. of Errors .
62 Errors .
62 .
Errors .
2. .
Meas .
-_ ________132 I27424 145 21025 14 I96 i8 324 31 96I I59 25282 I858'572 32 48 6 35 1225 i6 256 20 100 *712 3 254 6 30 900 9 8i II *82I 64 27 10 66 4356 II 122 77 5929 ^857 64 13 20 47 2209 22 484 22 441 *857 6o 25 I0 22 442 38 I444 32 2024 '876 56 9 IO 77 5929 56 3236 42 I682 -876 50 43 20 I63 26569 28 784 243 20449 *876 63 0 20 89 7921 27 289 33 I089 -887 53 33 IO 35 1225 55 3025 5 25 '887 53 0 20 I7 289 45 2025 42 2764 e89o 56 $8 IO 7 49 39 2522 84 7056 *89o 63 2 20 21 44I 26 676 85 7225 272 2924_ _ 97 9409 89 7922 i2)o0o00i 65I 12 II2 63 3969 24 576 152 23204 __=5 49 54 29I6 54 292I6 54 29i6 1858 833 +22-0 0 I8 324 120 I4400 II5 23225 .
_ I202 40402 95 9025 6i 3722 I25 49 75 5625 203 io609 253 23409 Zero =450 15 46 2II6 78 6084 66 4356 97 9409 32 2024 67 4489 324O ' I6 219 24I6I 89 7922 53 2809 Mean 64 4096 35 1225 6 36 Position .
56 3236 49 2402 6i 3722 _____ __ 42 2682 72 5I84 37 2369 24 I96 86 7396 I9 362 73 5329 25 225 30 9oo 205 2I025 36 2z96 4 i6 42 i68J 8I 656I 28 784 32 1024 69 476 I 206 I2236 248 229041 5 25 75 5625 220 I2200 97 9409 22 IZI 202 20202 9 8i 37 1369 40 i6oo 46 2ii6 95 9025 73 5329 132 I7I6I 52 2704 2o3 xo6og9 65 4225 87 7569 _6 256 267375 I55633 -~~~~207528 155633 z67375 _____ S 6_ 630536 Log. of 2 iz = 2049 1 I8o Log. of 630536 = 5-7997099 Log. of IIx 2045P3 30 Constant =55634093 4 0945410 13631192 o-68 25596 Log. of -454936 I 65795o3 4'0945410 P22 ; 23 074 _______ P 41*8035 Constant -5'5634093 W '04 33 Method 2 .
C ) / Errors .
e2 .
3 2-4 47 32446 2.0 400 32 3 486 29 36I 3 2-350 39 1521 324 I 37 1369 3 24 38i6 -2 56 325 IIII 12 I 32-3 57 44 1936 324 54 30 goo 324 57 27 729 324 33 30 goo 323 57 6 36 12 ) 53 29 30 goo Mean=324 27 e2 = 9429 Log. of 9429 3 97447 Log. I2 I *079I8 Constant =3 53738 Log. II= I-04I 39 1-5 I I85 3 212057 0'75593 Log. 4549 &amp;c. = I65795 P2 32 ' 50 2-I2057 P= 5-70x W= 0308 Constant 3'53738 Methiod 3 .
z379 7 .
s5f2=CS 39974=S2 5 f=CS 80279 = S3 9 f2 -CS3 ii4ii6=S4 19f2= CS4 13454-=S5 29f2-CS5 172308 3 =S6 I9 f2 = CS( 8I96o=S7 l9 2= C7 576658 I05 f2= CX ( S1+S2+ &amp; c ' ) Log. of 576658 =5-7609I83 Log. of '4549 &amp;c. -I6579503 5-4188686 1 , og .
of I05 -2002I1893.f2 2498 r5 3-3976793 Method 3 ( continued ) .
bl=l2 5 28 bl-b=2i &amp; ( b , -b)2= 441 b2= 25 29 b2-b=zo ... ... ... ... 400 b3= Iz6 27 b3-b=38 ... , , , .
I444 b4= I26 I4 b4-b=25 .
625 b5=125 40 b5-b9 ... ... ... ... 8I b6= 125 20 b6 b-b29 ... ... ... ... 841 b7=i26 o b7-b=ii .
IzI b I25 49 3953 See Method 1 .
Log. of 3953 =35969268 Log. of *4549 &amp;c. i 65795o3 32 54877 I Log. of 6=m-x =0-7781513 e2=299'73 =24767258 '%/ { ( 6 X 299.73 ?
2498.5 ) ?
( 6 &gt; K 299 73+2498-5 ) +(IO &gt; K 299X73+2498-5 ) P=~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ I +(26X 29973+24985 ) + ( 30 X 29973+2-498'5 ) +I( X 299-73+2498.5 ) + ( 20 X t99w73 +z4z20 5 ) }6X 299-73 = I798'38 20 X 299 73-=59946 +2498'5 X 6=0-77815 +2498'5 X 20= 130103 4296 88 &amp; X=3 63315 8q93.I &amp; X 3-92907 Ist W 001396 =314500 4th W ='002355 3 37i96 2nd W =00I396 30 X 29973 =899 I9 +2498'5 X 30=I.47712 10 X 299973 = 2997 3 11490-4 &amp; X -40o6b34 +2498'5 X io= I1 5th W -00726 II3 4I678 5495'8 &amp; X 3-74003 __ 3rd WV =-*OO8zo 325997 6th W -*002355 / th W O0023 5 Method 3 ( continued ) .
ist 'OOI396 X of I =0 .
znd='OOI396 X ofO04285=v15497 3rd= ooz 820 __ 4th='0o2355 I '84503 =X of P2 5th = 0026ii 09225z= of P 6th ='oo2355 7th=-002355 P2 69',99 , ='O 142.8 8P 8"'366 W -'014288 WW= 01429 125 4+ a , -28 W , = I4 z8 X 14= 392 02-29 W2= 14 29X I4= 406 .
' .
Meanresult=125 49 a3=87 W3= i8 87X i8 =I566 Zero =s450 i 04=74 W4= 24 74 X 24=1776 5 =-+0 W,5= 26 40X 26=1040 Mean Angle 324O 6 ' a6-20 W6= 24 20 X 24= 480 a7=6o W7= 24 6oX 24 1440 144 144 ) 7100(49'3 576 1340 I296 440 432 Summary of Results .
Mean No. of Name of Star .
PosiMeaSum of a2 e2 .
f2 P2 , P. W. Me .
tionl .
sures .
E~~~~~ 796e/ 0 7X3 48 1:796 ... ... .6 6i8 6o 2721370 4 ... ... . .
D5 591 29 6i 5.53'97 7'35 219 2 6I 4 , .
3101 2026'7 117-7 io85 93X x596,2Com .
Ber .
238 38 I56 3,529,521 66.4i 8-5 I5 I Early observations 238 36,..545.6I 2336 22 I843-8 .
238 36,. .
87310 2278'0 559'85 23.66 z31 1596 continued .
238 30 2I5 937,659 ... ... D *

