Professional Documents
Culture Documents
OF AVERAGING DATA
LVlien a number of measurements are made of any quantity variatioils \vill he found.
The question is: \Vhat is tlie best represcnt3t:ve value for the quantity thus mea.wrctl :
and how shall the precision oi the iiieasiireiiiciits be stated? The arithmetic iiicaii of all
the readings is generally taken a s the hest value. T o tell soiiictliiiiK almut tlie Iwecision
of the final result any one of five measures of variation which arc tliscu.sctl i n hooks dealiiip
with this subject may be given. These measures of deviation arc':
fi = probable error
a = the average deviation (from the arithmetic nicaiil
u = the standard deviation
1/11 = the reciprocal of tlie modulus of precisicm
k / z v = the reciprocal of the "precision constant"
Of these precision indexes the standard deviation. u. is most easily computctl. For the
set of observed values .rl, .r2..x,, of equal weight. the u for a siiiqlc observation is given hy
Z(.r - .r)?
$1
-1
-:4
=
v I1
-(.I'
r
I Z(.r-.r)?
--(X--.i-)*
Il(11-1)
I,.r):
112
The ratios of these precision indexes to one another for a iiornial (or Gaussian)
distribution are :
p : t i : u : 1 11 : i: 'it' : : 0.376936 : 1 *\;
: \,-)
: 1.000 : \'r
or roughly as p : n : u : 1/11 : k i t ' : : 7 : 8 : 10 : 14 : 25
Most experimental data can be represented by an equation of sonic form. One (it' thc
recommended methods for determining the coefficients oi such equations is the use of a
least-squares solution. This means that an attempt is niadc t o find values icir the coefficients
such that the sum of the squares of the deviations oi tlie cxpcriniciital points irom the
resulting curve has the least possible value. Certain tahles arc of help in making such
solutions (Tables 16-26), and reference shouitl be made to books or pspers on this suhjcct
for their use.
Xri example of one niethod of finding tlic cocfiicients of such sclcctctl equations (based
on "Treatment of Experimental Data," by \Vorthing and Gefiner. published h>- \\-ilcy.
1933) follows.
P a r t 1,-Least
Let, Q,, Q,. . . QI. be the k adjusted, but initially unkno\vn, values of the lincarly related
quantities. Let S,,
S2.. . S,,
be I i (> 1:) measured values oi Q's or o i linear combinations
of two or more, Q's.
Let &, &. . . A , $ be the adjustments or corrections that m i s t be applied to tlic iiieasii~cd
s ' s to yield consistent least-squares values for the Q's. See below for a simple illustration.
As O / J S C ~ P * O ~ ~cqircitiorrs
OII
we have
................................
+.
+
+
+.
.......................................................
r k i ~ i l ( _+
) l r/:ir~ilQ,+ i i ~ i ~ ~ ; +.
l c ).. .i/:rl:ilch
l
-rk,sil =0
OF AVERAGING D A T A (continued)
T A B L E 16.-METHODS
38
+ + +
+ + +.
+ + +.
(3)
.....................................
[ktatl = klal + hzaz + k3a3+ ...knan
Solutions of equation (2) yield'the least-squares adjusted values of
Qi,
Qz...ex.
For unequally weighted values of X , that is wl, wz,. . .wnfor X , X z . . .Xn, the rrornral
equations become
+
+
[ w t a ~ a t l Q ~ [wtatbtlQ2
[ ~ i b c a < l Q i [wtDtbtlQZ
+ [zvtatctIQ3+.. . [ W C ~ ~ -~ ~[wcatXtl
I Q X =0
+ [ ~ , b t ~ i l Q+..
3 .[ ~ r b t l ~ ~ l [Q~ ~t b t X i=
l 0
(4)
.....................................................................
+.
~w~k~aclQ
[ wl l h ~ b t l Q z I W ~ ~ ~ C ..~twthtktlQkI Q ~
of which
[ w t a d = zphalal
[zv~acbtl
= walbl
IwtkcXtl = 0
(5)
............................................
[wtk+atl= wlklal + wIkza2+ w3ksar+. . .wnknan
The weights wl, m . . .w,, associated with the Xi, X Z . . . X , and with the successive observation equations are taken as inversely proportional to the squares of the probable
errors (or of the standard deviations) of the corresponding X's. It is customary to take
simple rounded numbers for the proportional values. A precise set of 28, 50, 41, and 78
may be rounded to 3, 5, 4, and 8.
As a simple application, consider the elevations of stations B, C, and D above A. Let
those elevations in order be Q1, Q2,and Q3. Let the quantities measured and the observed
elevations be such as to yield the following observation equations :
Qz - Q 3 -12 ft = A5
Qi - Q3 - 5 ft = A6
Th coefficients al, b ~ and
,
are obvious. Substitution
Solutions of equation ( 7 ) yield 91 ft, 174 ft, and 44 ft for the elevations of B, C, and D
above A.
P a r t 2.-Least-squares
+ bx,
to represent a series of
For equally weighted pairs of (x,y) of which the errors of measurement are associated
with the determinations of the y's
of which
T A B L E 16.-METHODS
OF A V E R A G I N G D A T A (concluded)
39
The probable errors of the a and the b of equation (8) are given by
For unequally weighted measurements of which the errors of measurement are associated with the determinations of the y's,
Z w l x i l Z w , y i - Z w ix & v i xt y i
a=
Z z w Z w l x l z - (Zzehxr ) Z
Where the erroi s of measurement are associated with the x-determination only, the corb'y can be obtained by merely
responding coefficients of an equation of the type x = a'
interchanging x and y in equation ( 8 ) .
Where the errors of measurement are associated with both the x - and the y- determinations, the expressions are complicated."
Worthing, A. G . , and Geffner, J., Trcatment of experimental data, p. 259, John Wiley and Sons,
New York, 1943. Used by permmion.
+ cxz + dx3 to
P a r t 3.-Least-squares
represent a
For the general case involving irregularly spaced x-values, the formulae for a, b, c , etc.,
are very complex." However, for the case of equally weighted observations with errors
of measurement associated entirely with the y-values in which succeeding x-values are
equally spaced, the mechanics of the computations for least-squares constants are very
greatly simplified, thanks to tables computed by Baily and by Cox and Matuschak.Ia The
procedure requires a change of the x-variable to yield a new X-variable with a zero-value
at the midpoint of the series. I n case of an even number of terms, the shift is given by
x-x
4x
X,= -
(11)
of which Ax is the even spacing between successive x-values; and, if the number of terms
is odd, the shift is given by
-
x o = x4x/2
--x
(12)
+ +
(set)
(cm)
3
6
9
12
15
18
12.0
20.6
33.7
51.1
72.9
99.1
_-
-5
-3
-1
1
3
+
+
+5
289.4
-X2Y
(cm)
300.0
185.4
33.7
51.1
656.1
2477.5
-__
3703.8
40
T A B L E 17.-SHOWING
T H E MAKE-UP O F T H E CONSTANTS O F T H E LEASTSQUARES EQUATION O F T H E T Y P E y = a
bx
cx2 dxS FOR EQUATIONS OF VARYING DEGREES W H E N T H E ABBREVIATED M E T H O D O F
BAILEY AND O F COX AND MATUSCHAK IS U S E D *
+ +
This method is applicable only when succeeding values of x have a common difference
and a r e equally weighted. T h e independent variable, changed if necessary, must have a
zero value at the midpoint of the series with succeeding values differing by unity if the
number of terms is odd and by two if even. Values for the various k's, as computed by
Cox and Matuschak, are to be found in Tables 14 and 20.
J'or
L U ES O F P =
TABLE 18.-VA
.I
.2
.3
.4
0.5
.6
.7
.8
.9
1.o
.1
.2
.3
.4
1.5
.6
.7
.8
.9
2.0
.I
.2
.3
.4
2.5
.6
.7
.8
.9
7
8
9
.07886 .09008 .lo128
.I8999 .20094 .21184
,29742 ,30788 ,31828
,39921 .40901 ,41874
,49375 .SO275 51167
.52050
,60386
.67780
.74210
.79691
.52924
.61168
,68467
.74800
.80188 ,80677
3 4 9 4 .56332 ,57162
.63459 ,64203 ,64938
.70468 .71116 .71754
.76514 ,77067 .77610
,81156 .81627 32089 ,82542
.84270
.88021
.91031
.93401
,95229
.84681 .85084
.a353 .88679
.91296 .91553
,93606 .93807
.95385 .95538
3.5478 .85865
188997 .89308
.91805 .92051
.94002 .94191
.95686 ,95830
.86977
90200
.92751
.94731
,96237
.86244
,89612
,92290
,94376
.95970
,86614
.89910
.92524
,94556
,96105
.96611 ,96728
,97635 .97721
.98379 .98441
.98909 .98952
39279 ,99309
.96841
.97804
.98500
.98994
.99338
.96952
.97884
.98558
.99035
,99366
.97059
.97962
.98613
.99374
.99392
,99532
.99702
,99814
.99886
.99931
.99572
,99728
,99831
.99897
.99938
99591
.99741
.99839
.99902
.99941
,99552
,99715
.99822
.99891
.99935
,97162 .97263
.98038 .98110
,98667 .98719
,99111 ,99147
.99418 ,99443
.87333
,90484
,92973
.94902
.96365
,87680
90761
,93190
.95067
.96490
,99673
,99795
,99874
.99924
,99955
.99688
,99805
,99880
.99928
.99957
99.59 ,99961 .99963 .99965 .99967 ,99969 .99971 .99972 .99974 .99975
99976 .99978 ,99979 .99980 .99981 .99982 ..99983 .99984 ,99985 .99986
9 9 8 7 .99987 ,99988 .99989 .99989 .99s50 . 9 W 1 ,99991 ,99992 .99992
.99992 .99993 ,99993 .99994 .99994 .99994 .99995 .99995 .99995 ,99996
99996 S9996 999% .99997 ,99997 .99997 .99997 ,99997 .99997 .99998
.99998
3.0
-
.99999
.99999 1.00000
v)
T A B L E 19.-VALUES
O F T H E CONSTANTS, k,, E N T E R I N G LEAST-SQUARES SOLU TION S, U SIN G T H E A B B R E V I A T E D
M E T H O D O F B A I L Y A N D O F COX A N D M A T U S C H A K , W H E N T H E N U M B E R OF TER MS, n, IS O D D *
z
D
Q
T h e numbers in parentheses show the negative powers of 10 by which the adjacent numbers must he multiplied in order to obtain appropriate 12"'s.
ka
4
I
Im
(D
3
- 5
n 7
9
11
k4
kx
kG
3333 3333(8)
2000 0000
1428 5714
1111 1111
9090 9091(9)
so00 OOOO(8)
1000 0000
3571 4286(9)
1666 6667' '
9090 9091(10)
1000 OOOO(7)
4857 1429(8)
3333 3333
2554 1126
2074 5921
1000 0000(7',
1428 5714(6,
4761 9048(9)
2164 5022
1165 5012
1500 OOOO(7)
7142 8571(9)
1190 4762
3246 7532(10)
1165 5012
9027 7778(8)
2625 6614
1143 3782
6037 9435(9)
2361
3240
8277
2881
1111(8)
7407(9)
2166(10)
3779
6944 4444(9)
4629 6296(10)
7014 5903(11)
1618 7516
13
15
17
19
21
7692
6666
5882
5263
4761
5494 5055
3571 4286
2450 9803
1754 3860
1298 7013
1748 2517
1511 3122
1331 2693
1189 7391
1075 5149
6993 0070(10)
4524 8869
3095 9752
2211 4109
1634 5211
4995 0050(11)
2424 0465
1289 9897
7371 3696(12)
4457 7848
3584 6098
2304 5899
1570 2041
1118 3168
8248 9 7 0 ( 10)
1214 0637
5830 6799 ( 11)
3081 6420
1752 5617
1056 2015
4856 2549(12)
1745 7125
7166 6093(13)
3257 5497
1605 1694
23
25
27
29
31
4347 8261
4000 0000
3703 7037
3448 2759
3225 8065
9881
7692
6105
4926
4032
9813
9024
8352
7774
7270
6646(9)
1546
4904
0700
7048
1242 2360
%51 8357(11)
7662 8352
6179 7058
5056 1230
2823 2637
1858 0453
1263 1047
8828 1512(13)
6320 1537
6259
4862
3852
3104
2538
6672
4382
2974
2076
1485
8445
4692
2728
1650
1032
33
35
37
39
41
3030
2857
2702
2564
2439
3342 2460
2801 1204
2370 7918
2024 2915
1742 1603
6828
6437
6088
5775
5493
6552
3464
5061
5692
2589
4189
3510
2970
2535
2181
4620 6166
3441 1799
2605 2658
2001 6066
1558 2829
2102 4471
1760 7811
1489 3734
1271 0408
1093 4097
1084 7991
8073 4407(13)
6108 7522
4691 0081
3650 4910
6655 2091(15)
4402 0942
2979 8791
2059 2661
1449 7581
43
45
47
49
51
-
2325 5814
2222 2222
2127 6596
2040 8163
1960 7843
1510 1178
1317 5231
1156 3367
1020 4082
9049 7738(12)
5237
5004
4790
4595
4414
2849
1234
8525
0295
5960
1890 7166
1649 3485
1447 3875
1277 1066
1132 5285
1227 7380
9778 7451 (14)
7866 2362
6385 5329
5227 0545
9474
8263
7250
6396
5671
2875 1015
2289 2527
1841 0171
1494 1103
1222 7830
1037 9428
7545 3288( 16)
5561 9852
4152 6134
3136 9497
3077
6667
3529
1579
9048
3030
1429
7027
1026
0244
4229( 11)
3077
0061
1084
2581
3590
0035
0030
3684
5961
0791
3545
7423
7316
6983
1490(11)
1159
1033
2170
3855
0719(12)
3595
5336
4076
0296
6606(14)
0337
9299
5625
7049
3
!
T A B L E 20.-VALUES
OF T H E CONS T ANT S , k,, E N T E R I N G LEA ST-SQU A R ES SOLU TION S, U S I N G T H E A B B R E V I A T E D
M E T H O D O F B A I L Y A N D OF COX A N D M A T U S C H A K , W H E N T H E N U M B E R O F T E R M S , n, IS E V E N *
z
z
P
-0
The numbers in parentheses show the negative powers of 10 by which the adjacent numbers must be multiplied in order to obtain appropriate kn's.
b
r01
Im n
4
6
8
10
2500 OOOO(8)
1666 6667
1250 0000
1000 0000
5000
1428
5952
3030
0000(9)
5714
3810(10)
3030
6406
3945
2890
2289
12
14
16
18
20
8333 3333(9)
7142 8571
6250 0000
5555 5556
5000 0000
1748
1098
7352
5159
3759
2517
9011
9412(11)
9587
3985
22
24
26
28
30
4545 4545
4166 6667
3846 1538
3571 4286
3333 3333
32
34
36
38
40
3125
2941
2777
2631
2500
42
44
2380
2272
2173
2083
2000
-4
46
48
50
-
kz
k3
ks
k4
7812
1953
7812
3906
5030(9)
1250
5000(10)
2500
1562
1674
3720
1183
1897 3214
1621 0938
1415 5506
1256 5104
1129 7349
2232
1395
9300
6510
4734
1429
0893
5952(
4167
8485
4682 8172(12)
2146 2912
1094 1877
6046 8266( 13)
3560 0365
2823 2637
2173 9130
1709 4017
1368 3634
1112 3471
1026
9402
8675
8052
7513
2784
3164(9)
3091
8846
9509
3551 1364
2731 6434
2146 2912
1717 0330
1395 0893
2205
1425
9539
6578
4655
OOOO
1765
7778
5789
0000
9164
7634
6435
5471
4690
2229( 12)
4194
0064
0581
4315
7042
6627
6258
5927
5630
7390
2213
0624
9058
8741
1148
9574
8062
6853
5774
8971
1423(12)
4358
0703
0602
3369 1996
2486 7902
1867 7458
1424 7547
1102 0751
9524
7273
9130
3333
OOOO
4051
3523
3083
2713
2400
5355
6081
5646
8515
9604
5362
5118
4895
4690
4503
2160
0477
1643
8968
0048
5073
4411
3859
3396
3004
0520
3495
9309
7392
8077
8632
6839
5475
4424
3607
2500(8)
3125
6250
0625
SOOO(9)
1071(10)
2381(11)
7121
6748
2052
0720(14)
6704
4704
5332 ( 15)
3016
0792
7580
2121
7118 055619)
4870 756iiio)
9732 7441(11)
2964 3389
8683
2411
2630
5058
5556( 10)
2654(11)
4714( 12)
5988(13)
1149 4485
7125 6741(10)
4725 9399
1146 6157
5186 5517(
2622 0143
1440 7871
8448 3844( 3)
1348
4463
1722
7465
3540
9597
4695(14)
7426
2181(15)
8149
1790
1375
1079
8629
7006
5616
4794
5940
5508( 11)
8080
5218 8071
3364 5781
2248 0302
1548 2276
1094 4042
1805 8156
9775 0702( 16)
5561 6779
5767
4803
4043
3436
2944
1532
7846
7597
0952
4203
7913
5836
4380
3339
2582
1009(14)
2361
6481
8722
2837
1290 8811
8431 4304(17)
5643 7105
3861 1239
2693 8074
2542 3116
2210 2564
1933 6316
1701 3314
1504 8177
2021
1601
1281
1035
8439
9092
3580
5606
4426
3542(15)
1912 8753
1380 2431
1010 5351
7497 7742( 18)
5631 4922
6336
1126
4196
2040
8056(8)
7499
3534(9)
1329
jjoi iiig
2031 9424
T A B L E 21.-VALUES
e=
log e
1.0157
.0317
.0645
.lo52
.1175
0.00679
.01357
.02714
.04343
.04825
0.98450
.%923
.93941
.SO484
39484
1/3
1/2
3/!
1.1331
.1536
.1814
.2214
.2840
0.05429
.06204
.07238
.08686
.lo857
0.88250
.86688
.84648
.81873
.77880
3!2
7/;
1/64
1/32
1/16
1/10
1/9
:$
1/6
1/5
1/4
43
O F ex A N D e- A N D T H E I R L O G A R I T H M S
e-
9/4
5/2
e-=
1.3956
.6487
2.1170
.7183
3.4903
0.14476
.21715
.32572
.43429
.54287
0.71653
.60653
.47237
.36788
.28650
4.4817
5.7546
7.3891
9.4877
12.1825
0.65144
.76002
.86859
.97716
1.08574
0.22313
.17377
.13534
.lo540
.08208
5/4
T A B L E 22,-FURTHER
log @
e*
V A L U E S OF P
This table gives the values of the probability P, as defined in Table 18, corresponding to
different values of X / Y where r is the probable error. The probable error Y is equal to
0.47694/12.
0.0
.OooOo
.05378
.lo731
.16035
.21268
.01076
.06451
.11796
.17088
.22304
.01614
.06987
.12328
.17614
.22821
.02 152
.07523
.12860
.18138
.23336
.02690
.08059
.13391
.18662
.23851
.03228
.08594
.13921
.19185
.24364
.03766
.09129
.14451
.19707
.24876
.04303
.09663
.14980
.20229
.25388
.04840
0.1
0.2
0.3
0.4
.00538
,05914
.11264
.16562
.21787
.15508
.20749
.25898
0.5
.26407
.31430
.36317
.41052
.45618
.26915
.31925
.36798
.41517
.46064
.27421
.32419
.37277
.41979
.46509
.27927
.32911
.37755
.42440
.46952
.28431
.33402
.38231
.42899
.47393
.28934
.33892
.38705
.43357
.47832
.29436
.34380
.39178
.43813
.48270
.29936
.34866
.39649
.44267
.48705
.30435
,35352
.40118
,44719
.49139
.30933
.35835
.40586
.45169
.49570
1.o
.50000
.54188
.58171
.61942
165498
.50428
.54595
.58558
.62308
.65841
so853
.55001
.58912
.62671
.66182
.51277
55404
.59325
.63032
.66521
.51699
.55806
.59705
.63391
.66858
.52119
.56205
.60083
.63747
.67193
,52537
.56602
.60460
.64102
.67526
,52952
.56998
.60833
.64454
.67856
.53366
.57391
.61205
.64804
.68184
.53778
.57782
.61575
.65152
.68510
1.5
.68833
.71949
.74847
.77528
.79999
.69155
.72249
.75124
.77785
30235
.69474
.72546
.75400
.78039
30469
.69791
.72841
.75674
.78291
30700
.70106
.73134
.75945
.78542
.80930
,70419
,73425
.76214
.78790
.81158
.70729
.73714
.76481
.79036
31383
.71038
.74000
.76746
.79280
.81607
.71344
.74285
.77009
,79522
.81828
.71648
.74567
.77270
.79761
.82048
,82266
.84335
.86216
.87918
29450
32481
.84531
.86394
88078
.89595
32695
34726
A6570
,88237
,89738
32907
.84919
.86745
38395
.89879
.83117
.85109
.86917
,88550
.90019
A3324
A5298
37088
,88705
.90157
,83530
.85486
.87258
,88857
.90293
33734
35671
.87425
39008
.90428
.83936
.85854
.87591
39157
.90562
.84137
.86036
.87755
.89304
.90694
,90825
92051
.93141
.94105
,94954
.90954
.92166
.93243
.94195
.95033
.91082
.92280
.93344
.94284
.95111
.91208
.92392
.93443
.94371
.95187
.91332
.92503
,93541
.94458
.95263
.91456
,92613
.93638
.94543
,95338
.91578
.92721
.93734
.94627
.95412
.91698
.92828
.93828
.94711
.95484
.91817
,92934
.93922
.94793
.95557
.91935
,93038
.94014
.94874
.95628
0.6
0.7
0.8
0.9
1.1
1.2
1.3
1.4
1.6
1.7
1.8
1.9
2.0
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
3
50197
44
TABLE 23.-VALUES
OF T H E FACTOR 0.67454=
d..s
-
10
20
30
40
0.2248 0.2133
.1547 .1508
.1252 .1231
.lo80 .lo66
50
00
60
70
80
90
TABLE 24.-VALUES
OF T H E F A C T O R 0.6745
40
50
60
70
80
90
___
1) for the probable error of the
00
10
20
30
0.0711 0.0643
.0346 .0329
.0229 .0221
.0171 .0167
T A B L E 25.-LEAST
45
SQUARES
Part 1.-Values
zlvl
for the probable
= 0.8153 ___
d n ( n - 1)
11
10
20
30
40
0.0891 0.0806
.0434 .0412
.0287 .0277
.0214 .0209
50
00
60
70
80
90
---===
ndn -1
This factor occurs in the approximate equation yo = 0.8453 z(v( for the probable
Part 2.-Values
of 0.8453
ndgi
I1
-1
7
10
20
30
40
0.0282
.0097
.0052
,0034
50
00
60
70
80
90