The book "Statistics: a guide to the use of statistical methods in the Physical Sciences" is available from all good booksellers (and from Dillons) and can be ordered as ISBN 0 471 92295 1, for a very reasonable price.

A few mistakes have been found. News of any more will be gratefully received.

When the book was reprinted, somebody decided to redraw the front cover. Evil publishing gremlins introduced a couple of mistakes. The formula for the slope m reads 'xy-bar - xy-bar' instead of 'xy-bar - x-bar y-bar', and the formula for the constant c reads 'y-bar - m x' instead of 'ybar - m x-bar' (This one is 100% not my fault - the publisher's never even told me they were doing a reprint, let alone asked me to proofread what they'd done. If you've got one of the older copies, hang on to it!)

p5 Figure 2.2 In the central of the 3 figures, the y-axis labels for 10 and 20 are one tick-mark too low.

p25 In the un-numbered formula, (n-r!) should be (n-r)!

p29 In the example, prediction for the number of fatalities' should be prediction for the number of cases where no fatality occurred'

p33 in the second equation \lambda r' should be \lambda^r'

p41 Although the indexes of x, A etc are carefully introduced as having brackets: x(1) etc , these brackets are then omitted.

p42 The line after the xUUAUUX equation should start: This is \tilde x' A' x'

p46 The Cauchy function is also known as the Lorentzian.

p52 Line 5: commas can be inserted after 'and' and 'independent' if so desired.

p53 line 1 'them' should be 'then'. Line 2 'fall' should be 'falls'.

p57 Eq. 4.11 'cos(x,y)' should be 'cov(x,y)'

p57 line 8 'or errors' should be 'of errors'.

pp59-60 same comment as for p41

p60 The LHS of equation 4.18 is cov(f_k, f_l) not cov(f_k,f_i)

p62 line 10, p65 line 7, p66 line 1: for 'erros' read 'errors'

p67 Problem 4.5 'a current of 45 V' should be 'a voltage of 45 V'

p73 In the example: the variance of the midrange is $W^2 /2(N+1)(N+2)$; of the mean it is $W^2/12 N$. The refernce to 3.24 should be to 3.25 The last line should read (provided $N$ is greater than 2.)'

p 78 Eq 5.18 This should be made clearer. For a Gaussian distribution the variance is $$V(\hat V(x))=2 \sigma^4 (N-1)/N^2 \approx 2 \sigma^4/N$$ If Bessel's correction is applied, i.e. if Equation 5.14 is used instead of 5.13, this is multipled by $(N/(N-1))^2$ to give (for all N)
$$V(\hat V(x))= 2 \sigma^4/(N-1)$$

p78 The equation at the bottom is one equation not two -- the gap in the middle shouldn't be there

p80 Eq 5.25b should have an N in the numerator \hat{rho} = r = N (\bar{xy} - \bar{x} \bar{y}) /((N-1) s_x s_y)

p83 The second equation (for hat t) can be simplified a lot.

p85 The first expression in the third equation should be V( d ln L ...) not V< d ln L ...>

p 88 In Equation 5,35 the first term is cov^{-1} hat a_i, \hat a_j)

p89 In the equations at the top of the page, L shoud be replaced by ln L (5 times)

p89 Hats are also needed for mu and sigma on the left hand sides of the second set of equations

p94 bottom line mu is missing a subscript 2

p100 line 11 the formula for rho(m-hat,c-hat) is - xbar over sqrt(x-squared-bar), not x-bar-squared!

p100 line 13 There is a factor of N missing from the formula for chi**2

p101 Section 6.2.2 the x-squared in the denominator should be x-bar-squared

p102 Eq 6.12 Bars have been missed off the denominator, which should read x-squared-bar munus -x-bar-squared (not x-squared minus x-squared!) This error survived the corrections made in 1997.

p102 Eq 6.13 There is a factor of N missing from the formula for chi**2 p

p104 bar missing from x in denominator in first and second equations

p115 Eq 6.27 The bottom right element of the matrix is x-squared-bar minus x-bar-squared, not x-squared-bar minus x-squared-bar

p115 after Eq. 6.27

$$\Delta = \overline {x^2}\, \overline {x^4} - {\overline {x^3}}^2 - \overline {x}^2 \overline {x^4} + 2 \overline x \, \overline {x^2} \, \overline {x^3} - \overline {x^2}^3$$

p121 last line 'from a peak' should be 'from a pack'

p126 line 16 'three conventional way' should be 'three conventional ways'

p129 line 15 keV/C**2 should be keV/c**2

p129 bottom line The integration is over dx' not dx

p131 Eq 7.7 in the numerator 2 sigma 2 should be 2 sigma**2

and the x>0 should be X>0

p136 line 8 'Guassian' should be 'Gaussian'

p139 The Gamma (n+1/2) in the expression for the t distribution should be (n+1)/2

p150 Eq 8.5 and what follows: The term P(chi**2;n) is confusingly used for both the probability distribution for chi**2, as shown in Figure 6.4, and as the Integrated probability. In principle the distinction is always clear cut, but I admit the confusion isn't helpful! It would be better to have written the left hand side of Equation 8.5 as Prob(chi**2;N). Then Prob() should replace P() in the penultimate line of that page, and the 3rd line of page 151.

p154 In Equation 8.8, the subscripts on the combination terms are (r-3)/2 and (r-1)/2, not r-3/2 and r-1/2.

p161 the example should say row 6 and column 11

p170 line 4 '0.25 = 0.55' should be '0.25 × 0.55'

p183 The quadratic solutions should have q=- 1/2 etc

p188 Jay Orear's note has been reissued as report CLNS 82/511 (Cornell, 1982)

p191 Answer 4.6. The final line should read `breaks down, and so no sensible error can be quoted.''

Thanks to: Ross Barnett, Chris Collins, Andreas Dieckmann, Roman Gabrovsek, Mike Jones, Uli Katz, Steve MacMahon, Rastin Matin, Andrew McDonald, Arie Melamed-Katz, Karsten Riisager, James Weatherall, Juergen Wendland and others.