13.8 Spectral Analysis of Unevenly Sampled Data 575 In fact,since the MEM estimate may have very sharp spectral features,one wants to be able to evaluate it on a very fine mesh near to those features,but perhaps only more coarsely farther away from them.Here is a function which,given the coefficients already computed,evaluates (13.7.4)and returns the estimated power spectrum as a function of fA(the frequency times the sampling interval).Of course,fA should lie in the Nyquist range between-1/2 and 1/2. #include <math.h> float evlmem(float fdt,float d[],int m,float xms) Given d[1..m],m,xms as returned by memcof,this function returns the power spectrum estimate P(f)as a function of fdt =fA. f int i; float sumr=1.0,sumi=0.0; double wr=1.0,wi=0.0,wpr,wpi,wtemp,theta; Trig.recurrences in double precision. nted for thetas=6.28318530717959*fdt wpr=cos(theta); Set up for recurrence relations. wpi=sin(theta); for(i=1;1<=m;i+){ Loop over the terms in the sum wr=(wtemp=wr)*wpr-wi*wpi; RECIPES wi=wi*wpr+wtemp*wpi; sumr -d[i]*wr; These accumulate the denominator of(13.7.4). sumi -d[i]*wi; return xms/(sumr*sumr+sumi*sumi); Equation (13.7.4). 子二 Press. Be sure to evaluate P(f)on a fine enough grid to find any narrow features that may be there!Such narrow features,if present,can contain virtually all of the power in the data. You might also wish to know how the P(f)produced by the routines memcof and evlmem is normalized with respect to the mean square value of the input data vector.The answer is SCIENTIFIC r/ r1/2 6 P(fA)d(fA)=2 P(f△)d(f△)=mean square value of data (13.7.8) -1/2 n Sample spectra produced by the routines memcof and evlmem are shown in Figure 13.7.1. CITED REFERENCES AND FURTHER READING: Childers,D.G.(ed.)1978,Modern Spectrum Analysis(New York:IEEE Press),Chapter Il. Numerica 10621 Kay.S.M.,and Marple,S.L.1981,Proceedings of the /EEE,vol.69,pp.1380-1419. uctio Recipes 43108 (outside Software. 13.8 Spectral Analysis of Unevenly Sampled ying of Data Thus far,we have been dealing exclusively with evenly sampled data, hm=h(n△)n=.,-3,-2,-1,0,1,2,3,. (13.8.1) where A is the sampling interval,whose reciprocal is the sampling rate.Recall also (12.1) the significance of the Nyquist critical frequency 1 f:=2公 (13.8.2)
13.8 Spectral Analysis of Unevenly Sampled Data 575 Permission is granted for internet users to make one paper copy for their own personal use. Further reproduction, or any copyin Copyright (C) 1988-1992 by Cambridge University Press. Programs Copyright (C) 1988-1992 by Numerical Recipes Software. Sample page from NUMERICAL RECIPES IN C: THE ART OF SCIENTIFIC COMPUTING (ISBN 0-521-43108-5) g of machinereadable files (including this one) to any server computer, is strictly prohibited. To order Numerical Recipes books or CDROMs, visit website http://www.nr.com or call 1-800-872-7423 (North America only), or send email to directcustserv@cambridge.org (outside North America). In fact, since the MEM estimate may have very sharp spectral features, one wants to be able to evaluate it on a very fine mesh near to those features, but perhaps only more coarsely farther away from them. Here is a function which, given the coefficients already computed, evaluates (13.7.4) and returns the estimated power spectrum as a function of f∆ (the frequency times the sampling interval). Of course, f∆ should lie in the Nyquist range between −1/2 and 1/2. #include <math.h> float evlmem(float fdt, float d[], int m, float xms) Given d[1..m], m, xms as returned by memcof, this function returns the power spectrum estimate P(f) as a function of fdt = f∆. { int i; float sumr=1.0,sumi=0.0; double wr=1.0,wi=0.0,wpr,wpi,wtemp,theta; Trig. recurrences in double precision. theta=6.28318530717959*fdt; wpr=cos(theta); Set up for recurrence relations. wpi=sin(theta); for (i=1;i<=m;i++) { Loop over the terms in the sum. wr=(wtemp=wr)*wpr-wi*wpi; wi=wi*wpr+wtemp*wpi; sumr -= d[i]*wr; These accumulate the denominator of (13.7.4). sumi -= d[i]*wi; } return xms/(sumr*sumr+sumi*sumi); Equation (13.7.4). } Be sure to evaluate P(f) on a fine enough grid to find any narrow features that may be there! Such narrow features, if present, can contain virtually all of the power in the data. You might also wish to know how the P(f) produced by the routines memcof and evlmem is normalized with respect to the mean square value of the input data vector. The answer is 1/2 −1/2 P(f∆)d(f∆) = 2 1/2 0 P(f∆)d(f∆) = mean square value of data (13.7.8) Sample spectra produced by the routines memcof and evlmem are shown in Figure 13.7.1. CITED REFERENCES AND FURTHER READING: Childers, D.G. (ed.) 1978, Modern Spectrum Analysis (New York: IEEE Press), Chapter II. Kay, S.M., and Marple, S.L. 1981, Proceedings of the IEEE, vol. 69, pp. 1380–1419. 13.8 Spectral Analysis of Unevenly Sampled Data Thus far, we have been dealing exclusively with evenly sampled data, hn = h(n∆) n = ..., −3, −2, −1, 0, 1, 2, 3,... (13.8.1) where ∆ is the sampling interval, whose reciprocal is the sampling rate. Recall also (§12.1) the significance of the Nyquist critical frequency fc ≡ 1 2∆ (13.8.2)
576 Chapter 13.Fourier and Spectral Applications 1000 100 10 .com or call 1-800-872- (including this one) granted for interet -7423(North America to any server computer,is tusers to make one paper from NUMERICAL RECIPES IN C: 1988-1992 by Cambridge University Press.Programs THE 是 .15 2 .25 frequency Figure 13.7.1. Sample output of maximum entropy spectral estimation.The input signal consists of 512 samples of the sum of two sinusoids of very nearly the same frequency,plus white noise with about equal power.Shown is an expanded portion of the full Nyquist frequency interval (which would extend from zero to 0.5).The dashed spectral estimate uses 20 poles;the dotted,40;the solid,150.With the to dir Copyright (C) larger number of poles,the method can resolve the distinct sinusoids;but the flat noise background is beginning to show spurious peaks.(Note logarithmic scale.) ectcustser as codified by the sampling theorem:A sampled data set like equation (13.8.1)contains ART OF SCIENTIFIC COMPUTING(ISBN 0-521 complete information about all spectral components in a signal h(t)up to the Nyquist v@cam frequency,and scrambled or aliased information about any signal components at frequencies larger than the Nyquist frequency.The sampling theorem thus defines both the attractiveness, and the limitation,of any analysis of an evenly spaced data set. .Further reproduction, 1988-1992 by Numerical Recipes There are situations,however,where evenly spaced data cannot be obtained.A common -43108-5 case is where instrumental drop-outs occur,so that data is obtained only on a (not consecutive integer)subset of equation (13.8.1),the so-called missing data problem.Another case, (outside common in observational sciences like astronomy,is that the observer cannot completely control the time of the observations,but must simply accept a certain dictated set of ti's. North Software. There are some obvious ways to get from unevenly spaced t's to evenly spaced ones,as in equation (13.8.1).Interpolation is one way:lay down a grid of evenly spaced times on your Ame data and interpolate values onto that grid;then use FFT methods.In the missing data problem, you only have to interpolate on missing data points.If a lot of consecutive points are missing. you might as well just set them to zero,or perhaps"clamp"the value at the last measured point. However,the experience of practitioners of such interpolation techniques is not reassuring Generally speaking,such techniques perform poorly.Long gaps in the data,for example, often produce a spurious bulge of power at low frequencies(wavelengths comparable to gaps) A completely different method of spectral analysis for unevenly sampled data,one that mitigates these difficulties and has some other very desirable properties,was developed by Lomb[1],based in part on earlier work by Barning [2]and Vanicek [3],and additionally elaborated by Scargle [4].The Lomb method (as we will call it)evaluates data,and sines
576 Chapter 13. Fourier and Spectral Applications Permission is granted for internet users to make one paper copy for their own personal use. Further reproduction, or any copyin Copyright (C) 1988-1992 by Cambridge University Press. Programs Copyright (C) 1988-1992 by Numerical Recipes Software. Sample page from NUMERICAL RECIPES IN C: THE ART OF SCIENTIFIC COMPUTING (ISBN 0-521-43108-5) g of machinereadable files (including this one) to any server computer, is strictly prohibited. To order Numerical Recipes books or CDROMs, visit website http://www.nr.com or call 1-800-872-7423 (North America only), or send email to directcustserv@cambridge.org (outside North America). power spectral densitty 0.1 1 10 100 1000 .1 .15 .2 .25 .3 frequency f Figure 13.7.1. Sample output of maximum entropy spectral estimation. The input signal consists of 512 samples of the sum of two sinusoids of very nearly the same frequency, plus white noise with about equal power. Shown is an expanded portion of the full Nyquist frequency interval (which would extend from zero to 0.5). The dashed spectral estimate uses 20 poles; the dotted, 40; the solid, 150. With the larger number of poles, the method can resolve the distinct sinusoids; but the flat noise background is beginning to show spurious peaks. (Note logarithmic scale.) as codified by the sampling theorem: A sampled data set like equation (13.8.1) contains complete information about all spectral components in a signal h(t) up to the Nyquist frequency, and scrambled or aliased information about any signal components at frequencies larger than the Nyquist frequency. The sampling theorem thus defines both the attractiveness, and the limitation, of any analysis of an evenly spaced data set. There are situations, however, where evenly spaced data cannot be obtained. A common case is where instrumental drop-outs occur, so that data is obtained only on a (not consecutive integer) subset of equation (13.8.1), the so-called missing data problem. Another case, common in observational sciences like astronomy, is that the observer cannot completely control the time of the observations, but must simply accept a certain dictated set of ti’s. There are some obvious ways to get from unevenly spaced ti’s to evenly spaced ones, as in equation (13.8.1). Interpolation is one way: lay down a grid of evenly spaced times on your data and interpolate values onto that grid; then use FFT methods. In the missing data problem, you only have to interpolate on missing data points. If a lot of consecutive points are missing, you might as well just set them to zero, or perhaps “clamp” the value at the last measured point. However, the experience of practitioners of such interpolation techniques is not reassuring. Generally speaking, such techniques perform poorly. Long gaps in the data, for example, often produce a spurious bulge of power at low frequencies (wavelengths comparable to gaps). A completely different method of spectral analysis for unevenly sampled data, one that mitigates these difficulties and has some other very desirable properties, was developed by Lomb [1], based in part on earlier work by Barning [2] and Van´ıcek ˇ [3], and additionally elaborated by Scargle [4]. The Lomb method (as we will call it) evaluates data, and sines
13.8 Spectral Analysis of Unevenly Sampled Data 577 and cosines,only at times t;that are actually measured.Suppose that there are N data pointshi=h(ti),i=1,...,N.Then first find the mean and variance of the data by the usual formulas. 1 =立a- (13.8.3) 1 Now,the Lomb normalized periodogram (spectral power as a function of angular frequency w=2f>0)is defined by Pw(u)≡ [区，仙-刀s6-，-列血刊 22 ∑c0s2w(4-T) ∑，sin2w(化，-T) (138.4) Here T is defined by the relation ∑jsin2wt tan(2WT)= (13.8.5) ∑，cos2w5 RECIPES The constant r is a kind of offset that makes P(w)completely independent of shifting 令 all the ti's by any constant.Lomb shows that this particular choice of offset has another, deeper,effect:It makes equation (13.8.4)identical to the equation that one would obtain if one estimated the harmonic content of a data set,at a given frequency w,by linear least-squares Press. fitting to the model h(t)=Acoswt+B sinwt (13.8.6) This fact gives some insight into why the method can give results superior to FFT methods:It weights the data on a"per point"basis instead of on a"per time interval"basis,when uneven sampling can render the latter seriously in error. SCIENTIFIC A very common occurrence is that the measured data points h are the sum of a periodic signal and independent (white)Gaussian noise.If we are trying to determine the presence 6 or absence of such a periodic signal,we want to be able to give a quantitative answer to the question,"How significant is a peak in the spectrum Py(w)?"In this question,the null hypothesis is that the data values are independent Gaussian random values.A very nice property of the Lomb normalized periodogram is that the viability of the null hypothesis can be tested fairly rigorously,as we now discuss. The word"normalized"refers to the factor o2 in the denominator of equation(13.8.4). 10621 Scargle [4]shows that with this normalization,at any particular w and in the case of the null Numerica hypothesis,PN(w)has an exponential probability distribution with unit mean.In other words, the probability that P(w)will be between some positive z and z+dz is exp(-z)dz.It uctio 43106 readily follows that,if we scan some M independent frequencies,the probability that none Recipes give values larger than z is (1-e-=)M.So P(>)≡1-(1-e)M (13.8.7) North Software. is the false-alarm probability of the null hypothesis,that is,the significance level of any peak in PN(w)that we do see.A small value for the false-alarm probability indicates a highly significant periodic signal. To evaluate this significance,we need to know M.After all,the more frequencies we look at,the less significant is some one modest bump in the spectrum.(Look long enough, find anything!)A typical procedure will be to plot Py(w)as a function of many closely spaced frequencies in some large frequency range.How many of these are independent? Before answering,let us first see how accurately we need to know M.The interesting region is where the significance is a small (significant)number,<1.There,equation (13.8.7) can be series expanded to give P(>z)≈Me- (13.8.8)
13.8 Spectral Analysis of Unevenly Sampled Data 577 Permission is granted for internet users to make one paper copy for their own personal use. Further reproduction, or any copyin Copyright (C) 1988-1992 by Cambridge University Press. Programs Copyright (C) 1988-1992 by Numerical Recipes Software. Sample page from NUMERICAL RECIPES IN C: THE ART OF SCIENTIFIC COMPUTING (ISBN 0-521-43108-5) g of machinereadable files (including this one) to any server computer, is strictly prohibited. To order Numerical Recipes books or CDROMs, visit website http://www.nr.com or call 1-800-872-7423 (North America only), or send email to directcustserv@cambridge.org (outside North America). and cosines, only at times ti that are actually measured. Suppose that there are N data points hi ≡ h(ti), i = 1,...,N. Then first find the mean and variance of the data by the usual formulas, h ≡ 1 N N 1 hi σ2 ≡ 1 N − 1 N 1 (hi − h) 2 (13.8.3) Now, the Lomb normalized periodogram (spectral power as a function of angular frequency ω ≡ 2πf > 0) is defined by PN (ω) ≡ 1 2σ2 j (hj − h) cos ω(tj − τ ) 2 j cos2 ω(tj − τ ) + j (hj − h) sin ω(tj − τ ) 2 j sin2 ω(tj − τ ) (13.8.4) Here τ is defined by the relation tan(2ωτ ) = j sin 2ωtj j cos 2ωtj (13.8.5) The constant τ is a kind of offset that makes PN (ω) completely independent of shifting all the ti’s by any constant. Lomb shows that this particular choice of offset has another, deeper, effect: It makes equation (13.8.4) identical to the equation that one would obtain if one estimated the harmonic content of a data set, at a given frequency ω, by linear least-squares fitting to the model h(t) = A cos ωt + B sin ωt (13.8.6) This fact gives some insight into why the method can give results superior to FFT methods: It weights the data on a “per point” basis instead of on a “per time interval” basis, when uneven sampling can render the latter seriously in error. A very common occurrence is that the measured data points hi are the sum of a periodic signal and independent (white) Gaussian noise. If we are trying to determine the presence or absence of such a periodic signal, we want to be able to give a quantitative answer to the question, “How significant is a peak in the spectrum PN (ω)?” In this question, the null hypothesis is that the data values are independent Gaussian random values. A very nice property of the Lomb normalized periodogram is that the viability of the null hypothesis can be tested fairly rigorously, as we now discuss. The word “normalized” refers to the factor σ2 in the denominator of equation (13.8.4). Scargle [4] shows that with this normalization, at any particular ω and in the case of the null hypothesis, PN (ω) has an exponential probability distribution with unit mean. In other words, the probability that PN (ω) will be between some positive z and z + dz is exp(−z)dz. It readily follows that, if we scan some M independent frequencies, the probability that none give values larger than z is (1 − e−z) M. So P(> z) ≡ 1 − (1 − e−z ) M (13.8.7) is the false-alarm probability of the null hypothesis, that is, the significance level of any peak in PN (ω) that we do see. A small value for the false-alarm probability indicates a highly significant periodic signal. To evaluate this significance, we need to know M. After all, the more frequencies we look at, the less significant is some one modest bump in the spectrum. (Look long enough, find anything!) A typical procedure will be to plot PN (ω) as a function of many closely spaced frequencies in some large frequency range. How many of these are independent? Before answering, let us first see how accurately we need to know M. The interesting region is where the significance is a small (significant) number, 1. There, equation (13.8.7) can be series expanded to give P(> z) ≈ Me−z (13.8.8)
578 Chapter 13.Fourier and Spectral Applications T 0 01 2 00 000 0 0 0 0 1 00 00 00 0 0 00 0 00 0 0 0 00 0 -1 00000 0 0 0 0 00 0c9 0 2 8 0 0 0 0 Permission is 0 10 20 30 40 50 60 70 80 90 100 83 time 14 11-600 (including this one) granted for ….001 .005 10 ------.01 8 .05 6 (North America to any server computer, tusers to make one paper 1988-1992 by Cambridge University Press. from NUMERICAL RECIPES IN C: THE 2 是 ART 0 .1 .2 3 4 .5 .6 .7 8 .9 1 Programs frequency send copy for their Figure 13.8.1. Example of the Lomb algorithm in action.The 100 data points(upper figure)are at random times between 0 and 100.Their sinusoidal component is readily uncovered (lower figure)by the algorithm,at a significance level better than 0.001.If the 100 data points had been evenly spaced at to dir Copyright(C) unit interval,the Nyquist critical frequency would have been 0.5.Note that,for these unevenly spaced points,there is no visible aliasing into the Nyquist range. ectcustser We see that the significance scales linearly with M.Practical significance levels are numbers 1988-1992 by Numerical Recipes OF SCIENTIFIC COMPUTING(ISBN 0-521- like 0.05,0.01,0.001,etc.An error of even +50%in the estimated significance is often v@cam tolerable,since quoted significance levels are typically spaced apart by factors of 5 or 10.So our estimate of M need not be very accurate. Horne and Baliunas [5]give results from extensive Monte Carlo experiments for deter- mining M in various cases.In general M depends on the number of frequencies sampled, -431085 the number of data points N,and their detailed spacing.It turns out that M is very nearly equal to N when the data points are approximately equally spaced,and when the sampled (outside frequencies "fill"(oversample)the frequency range from 0 to the Nyquist frequency fe (equation 13.8.2).Further,the value of M is not importantly different for random spacing of North Software. the data points than for equal spacing.When a larger frequency range than the Nyquist range is sampled,M increases proportionally.About the only case where M differs significantly from the case of evenly spaced points is when the points are closely clumped,say into groups of 3;then(as one would expect)the number of independent frequencies is reduced visit website by a factor of about 3. The program period,below,calculates an effective value for M based on the above rough-and-ready rules and assumes that there is no important clumping.This will be adequate for most purposes.In any particular case,if it really matters,it is not too difficult to compute a better value of M by simple Monte Carlo:Holding fixed the number of data points and their locations ti,generate synthetic data sets of Gaussian(normal)deviates,find the largest values of P(w)for each such data set(using the accompanying program),and fit the resulting distribution for M in equation (13.8.7)
578 Chapter 13. Fourier and Spectral Applications Permission is granted for internet users to make one paper copy for their own personal use. Further reproduction, or any copyin Copyright (C) 1988-1992 by Cambridge University Press. Programs Copyright (C) 1988-1992 by Numerical Recipes Software. Sample page from NUMERICAL RECIPES IN C: THE ART OF SCIENTIFIC COMPUTING (ISBN 0-521-43108-5) g of machinereadable files (including this one) to any server computer, is strictly prohibited. To order Numerical Recipes books or CDROMs, visit website http://www.nr.com or call 1-800-872-7423 (North America only), or send email to directcustserv@cambridge.org (outside North America). −2 −1 0 1 2 0 10 20 30 40 50 60 70 80 90 100 time amplitude .001 .005 .01 .05 .1 .5 0 2 4 6 8 10 12 14 power 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 frequency significance levels Figure 13.8.1. Example of the Lomb algorithm in action. The 100 data points (upper figure) are at random times between 0 and 100. Their sinusoidal component is readily uncovered (lower figure) by the algorithm, at a significance level better than 0.001. If the 100 data points had been evenly spaced at unit interval, the Nyquist critical frequency would have been 0.5. Note that, for these unevenly spaced points, there is no visible aliasing into the Nyquist range. We see that the significance scales linearly with M. Practical significance levels are numbers like 0.05, 0.01, 0.001, etc. An error of even ±50% in the estimated significance is often tolerable, since quoted significance levels are typically spaced apart by factors of 5 or 10. So our estimate of M need not be very accurate. Horne and Baliunas [5] give results from extensive Monte Carlo experiments for determining M in various cases. In general M depends on the number of frequencies sampled, the number of data points N, and their detailed spacing. It turns out that M is very nearly equal to N when the data points are approximately equally spaced, and when the sampled frequencies “fill” (oversample) the frequency range from 0 to the Nyquist frequency fc (equation 13.8.2). Further, the value of M is not importantly different for random spacing of the data points than for equal spacing. When a larger frequency range than the Nyquist range is sampled, M increases proportionally. About the only case where M differs significantly from the case of evenly spaced points is when the points are closely clumped, say into groups of 3; then (as one would expect) the number of independent frequencies is reduced by a factor of about 3. The program period, below, calculates an effective value for M based on the above rough-and-ready rules and assumes that there is no important clumping. This will be adequate for most purposes. In any particular case, if it really matters, it is not too difficult to compute a better value of M by simple Monte Carlo: Holding fixed the number of data points and their locations ti, generate synthetic data sets of Gaussian (normal) deviates, find the largest values of PN (ω) for each such data set (using the accompanying program), and fit the resulting distribution for M in equation (13.8.7)
13.8 Spectral Analysis of Unevenly Sampled Data 579 Figure 13.8.1 shows the results of applying the method as discussed so far.In the upper figure,the data points are plotted against time.Their number is N=100,and their distribution in t is Poisson random.There is certainly no sinusoidal signal evident to the eye. The lower figure plots PN(w)against frequency f=w/2.The Nyquist critical frequency that would obtain if the points were evenly spaced is at f fe=0.5.Since we have searched up to about twice that frequency,and oversampled the f's to the point where successive values of PN(w)vary smoothly,we take M 2N.The horizontal dashed and dotted lines are (respectively from bottom to top)significance levels 0.5,0.1,0.05,0.01,0.005,and 0.001. One sees a highly significant peak at a frequency of 0.81.That is in fact the frequency of the sine wave that is present in the data.(You will have to take our word for this!) Note that two other peaks approach,but do not exceed the 50%significance level;that is about what one might expect by chance.It is also worth commenting on the fact that the 8 significant peak was found (correctly)above the Nyquist frequency and without any significant aliasing down into the Nyquist interval!That would not be possible for evenly spaced data.It is possible here because the randomly spaced data has some points spaced much closer than ed for the "average"sampling rate,and these remove ambiguity from any aliasing. Implementation of the normalized periodogram in code is straightforward,with,however. a few points to be kept in mind.We are dealing with a slow algorithm.Typically,for N data points,we may wish to examine on the order of 2N or 4N frequencies.Each combination of frequency and data point has,in equations (13.8.4)and (13.8.5),not just a few adds or multiplies,but four calls to trigonometric functions;the operations count can easily reach several hundred times N2.It is highly desirable-in fact results in a factor 4 speedup- to replace these trigonometric calls by recurrences.That is possible only if the sequence of frequencies examined is a linear sequence.Since such a sequence is probably what most users would want anyway,we have built this into the implementation. 代 Press. ART At the end of this section we describe a way to evaluate equations (13.8.4)and (13.8.5) approximately,but to any desired degree of approximation-by a fast method [6]whose operation count goes only as N log N.This faster method should be used for long data sets. 9 Program The lowest independent frequency f to be examined is the inverse of the span of the input data,maxi(ti)-mini(ti)=T.This is the frequency such that the data can include one complete cycle.In subtracting off the data's mean,equation (13.8.4)already assumed that you are not interested in the data's zero-frequency piece-which is just that mean value.In an FFT method,higher independent frequencies would be integer multiples of 1/T.Because we to dir are interested in the statistical significance of any peak that may occur,however,we had better (over-)sample more finely than at interval 1/T,so that sample points lie close to the top of OF SCIENTIFIC COMPUTING (ISBN any peak.Thus,the accompanying program includes an oversampling parameter,called ofac; 1988-19920 a value ofac4 might be typical in use.We also want to specify how high in frequency to go,say fhi.One guide to choosing fhi is to compare it with the Nyquist frequency fe which would obtain if the N data points were evenly spaced over the same span T,that is 10621 fe=N/(2T).The accompanying program includes an input parameter hifac,defined as fhi/fe.The number of different frequencies Np returned by the program is then given by Numerical Recipes 43106 Np=ofac×hifac N (13.8.9) 2 (outsi (You have to remember to dimension the output arrays to at least this size.) The code does the trigonometric recurrences in double precision and embodies a few Software. tricks with trigonometric identities,to decrease roundoff errors.If you are an aficionado of such things you can puzzle it out.A final detail is that equation (13.8.7)will fail because of roundoff error if z is too large;but equation (13.8.8)is fine in this regime. #include <math.h> #include "nrutil.h" #def1neTW0PID6.2831853071795865 void period(float x[],float y[],int n,float ofac,float hifac,float px[], float py[,int np,int *nout,int *jmax,float *prob) Given n data points with abscissas x[1..n](which need not be equally spaced)and ordinates y[1..n],and given a desired oversampling factor ofac (a typical value being 4 or larger). this routine fills array px[1..np]with an increasing sequence of frequencies (not angular
13.8 Spectral Analysis of Unevenly Sampled Data 579 Permission is granted for internet users to make one paper copy for their own personal use. Further reproduction, or any copyin Copyright (C) 1988-1992 by Cambridge University Press. Programs Copyright (C) 1988-1992 by Numerical Recipes Software. Sample page from NUMERICAL RECIPES IN C: THE ART OF SCIENTIFIC COMPUTING (ISBN 0-521-43108-5) g of machinereadable files (including this one) to any server computer, is strictly prohibited. To order Numerical Recipes books or CDROMs, visit website http://www.nr.com or call 1-800-872-7423 (North America only), or send email to directcustserv@cambridge.org (outside North America). Figure 13.8.1 shows the results of applying the method as discussed so far. In the upper figure, the data points are plotted against time. Their number is N = 100, and their distribution in t is Poisson random. There is certainly no sinusoidal signal evident to the eye. The lower figure plots PN (ω) against frequency f = ω/2π. The Nyquist critical frequency that would obtain if the points were evenly spaced is at f = fc = 0.5. Since we have searched up to about twice that frequency, and oversampled the f’s to the point where successive values of PN (ω) vary smoothly, we take M = 2N. The horizontal dashed and dotted lines are (respectively from bottom to top) significance levels 0.5, 0.1, 0.05, 0.01, 0.005, and 0.001. One sees a highly significant peak at a frequency of 0.81. That is in fact the frequency of the sine wave that is present in the data. (You will have to take our word for this!) Note that two other peaks approach, but do not exceed the 50% significance level; that is about what one might expect by chance. It is also worth commenting on the fact that the significant peak was found (correctly) above the Nyquist frequency and without any significant aliasing down into the Nyquist interval! That would not be possible for evenly spaced data. It is possible here because the randomly spaced data has some points spaced much closer than the “average” sampling rate, and these remove ambiguity from any aliasing. Implementation of the normalized periodogram in code is straightforward, with, however, a few points to be kept in mind. We are dealing with a slow algorithm. Typically, for N data points, we may wish to examine on the order of 2N or 4N frequencies. Each combination of frequency and data point has, in equations (13.8.4) and (13.8.5), not just a few adds or multiplies, but four calls to trigonometric functions; the operations count can easily reach several hundred times N2. It is highly desirable — in fact results in a factor 4 speedup — to replace these trigonometric calls by recurrences. That is possible only if the sequence of frequencies examined is a linear sequence. Since such a sequence is probably what most users would want anyway, we have built this into the implementation. At the end of this section we describe a way to evaluate equations (13.8.4) and (13.8.5) — approximately, but to any desired degree of approximation — by a fast method [6] whose operation count goes only as N log N. This faster method should be used for long data sets. The lowest independent frequency f to be examined is the inverse of the span of the input data, maxi(ti)−mini(ti) ≡ T. This is the frequency such that the data can include one complete cycle. In subtracting off the data’s mean, equation (13.8.4) already assumed that you are not interested in the data’s zero-frequency piece — which is just that mean value. In an FFT method, higher independent frequencies would be integer multiples of 1/T. Because we are interested in the statistical significance of any peak that may occur, however, we had better (over-) sample more finely than at interval 1/T, so that sample points lie close to the top of any peak. Thus, the accompanying program includes an oversampling parameter, called ofac; a value ofac >∼ 4 might be typical in use. We also want to specify how high in frequency to go, say fhi. One guide to choosing fhi is to compare it with the Nyquist frequency fc which would obtain if the N data points were evenly spaced over the same span T, that is fc = N/(2T). The accompanying program includes an input parameter hifac, defined as fhi/fc. The number of different frequencies NP returned by the program is then given by NP = ofac × hifac 2 N (13.8.9) (You have to remember to dimension the output arrays to at least this size.) The code does the trigonometric recurrences in double precision and embodies a few tricks with trigonometric identities, to decrease roundoff errors. If you are an aficionado of such things you can puzzle it out. A final detail is that equation (13.8.7) will fail because of roundoff error if z is too large; but equation (13.8.8) is fine in this regime. #include <math.h> #include "nrutil.h" #define TWOPID 6.2831853071795865 void period(float x[], float y[], int n, float ofac, float hifac, float px[], float py[], int np, int *nout, int *jmax, float *prob) Given n data points with abscissas x[1..n] (which need not be equally spaced) and ordinates y[1..n], and given a desired oversampling factor ofac (a typical value being 4or larger), this routine fills array px[1..np] with an increasing sequence of frequencies (not angular