Aufgaben:Exercise 4.11Z: C Program "acf2"

• Return to book

C program  $2$  for ACF–calculation correction

You can see on the right the C program "acf2" for calculating the discrete ACF values  $\varphi_x(k)$  with index  $k = 0$, ... , $l$.

In contrast to the program "acf1" from  Exercise 4.11  here the algorithm described in the theory part is applied directly.  It should be noted:

• Let the long value passed to the program be here  $l=10$.
• The calculated ACF values  $\varphi_x(0)$, ... , $\varphi_x(10)$  are returned to the main program with the float field  $\rm ACF[ \ ]$  In lines 7 and 8, this field is prefilled with zeros.
• The random variable  $x( \ )$  is defined as a float function in line 4, as is an auxiliary field  ${\rm H}[10000 ]$, into which the  $N = 10000$  samples  $x_\nu$  are entered (lines 9 and 10).
• The names of the run variables in line 6 are adapted to the given algorithm.
• The actual ACF calculation is done from line 11.  This program part is marked red in the program code.

Hints:

• This exercise belongs to the chapter  Auto-Correlation Function.
• In particular, reference is made to the pages 
  • Numerical ACF determination,
  • Accuracy of the numerical ACF calculation.

Questions

1

On how many summands  ($S$)  is the ACF calculation based for index  $k=0$   resp.  for  $k=10$ ?

$S_{k=0} \ = \ $

$S_{k=10} \ = \ $

2

Which of the following statements are correct?

	The calculation time increases linearly with  $l + 1$, i.e. with the number of ACF values to be calculated.
	The calculation time increases quadratically with the number  $N$  of samples considered.
	The calculation becomes more accurate as  $N$  increases.
	If a float variable is represented with  $\rm 4 \ bytes$ , "acf2" needs at least  $4 \cdot N$  bytes of memory.

3

Which of the following statements are true?

	The stronger the internal statistical bindings of the process are, the more inexact is the ACF result for a given  $N$ .
	The stronger the internal statistical bindings of the process are, the more accurate, for a given  $N$  is the ACF result.
	If the process has statistical bindings, the errors of the numerical ACF calculation are also correlated.     Example:     If the value  $\varphi_x(k=5)$  is too large, it is very likely that  $\varphi_x(k=4)$  and  $\varphi_x(k=6)$  will also be too large.

Solution

Solution

(1)  To calculate the ACF value  $\varphi_x(0)$  is averaged over  $\underline{N =10000}$  summands, for  $\varphi_x(10)$  only over  $\underline{N = 9990}$.

(2)  Correct are the proposed solutions 1, 3 and 4:

• The computation time increases approximately linearly with  $N$  and  $l + 1$  as can be seen from the ACF calculation highlighted in red.
• In contrast, the computation time for the other parts of the program can be neglected.
• Naturally, the calculation also becomes more accurate with increasing  $N$ .
• This goes here – in contrast to the program "acf1" of exercise 4.11 – however at the expense of the necessary memory requirement.
• Since each float variable takes up exactly four bytes, the auxiliary field  ${\rm H}[10000 ]$  alone requires a memory of 40 kByte.

(3)  Correct are suggested solutions 1 and 3:

• The stronger the statistical bindings within the random process, the less accurate is the ACF calculation for a given  $N$ .
• This fact can be illustrated for example by the power calculation  $($ACF value at  $k=0)$ :
  • If all  $N$  samples are statistically independent, then all contributions provide the maximum information about the ACF–value  $\varphi_x(k=0)$.
  • However, if there are statistical bindings between  $x_\nu$  and  $x_{\nu+1}$, but not between  $x_\nu$  and  $x_{\nu+2}$, only half of all samples provide the full information  about  $\varphi_x(k=0)$  and the others provide only limited information.
• This loss of information based on correlations can only be compensated in this example by doubling  $N$ .
• The last statement is also true, as explained in detail on the page "Accuracy of the numerical ACF calculation" in the theory section.