NESC1057 PCC/SRC.

1. NAME OR DESIGNATION OF PROGRAM:  PCC/SRC.

3. DESCRIPTION OF PROGRAM OR FUNCTION

PCC/SRC is designed for use in  conjunction with sensitivity analyses of complex computer models. PCC/SRC calculates the partial correlation coefficients (PCC) and the standardized regression coefficients (SRC) from the multivariate input to, and output from, a computer model.

4. METHOD OF SOLUTION

PCC/SRC calculates the coefficients on either the original observations or on the ranks of the original observations. These coefficients provide alternative measures of the relative contribution (importance) of each of the various input variables to the observed variations in output. Relationships between the coefficients and differences in their interpretations are identified. If the computer model output has an associated time  or spatial history, PCC/SRC will generate a graph of the coefficients over time or space for each input-variable, output- variable combination of interest, indicating the importance of each  input value over time or space.

5. RESTRICTIONS ON THE COMPLEXITY OF THE PROBLEM

Maxima of -
   100 observations
100 different time steps or intervals between successive        dependent variable readings
    50 independent variables (model input)
    20 dependent variables (model output)
10 ordered triples specifying intervals between dependent        variable readings

6. TYPICAL RUNNING TIME

NESC executed the sample problem in 30 CPU seconds on a DEC VAX11/780.

The sample problem was executed at the NEA-DB on a DEC  VAX 6000-510 computer in about 14 seconds of CPU time.

7. UNUSUAL FEATURES OF THE PROGRAM:

8. RELATED AND AUXILIARY PROGRAMS

Plots generated by PCC/SRC are produced using the CA-DISSPLA proprietary software package Version 9.0 (see http://www.gaeinc.com). PCC/SRC is most useful in analysing input-output relationships  when the input has been selected using the Latin hypercube sample program LHS (NESC 1085).

10. REFERENCES

- R. L. Iman, and M.J. Shortencarier
  A FORTRAN 77 Program and User's Guide for the Generation of Latin
  Hypercube and Random Samples for Use with Computer Models,
  NUREG/CR-3624 (SAND83-2365), March 1984.

- R.L. Iman, M.J. Shortencarier, and J.D. Johnson:
  A FORTRAN 77 Program and User's Guide for the Calculation of
  Partial Correlation and Standardized Regression Coefficients
  NUREG/CR-4122 (SAND85-0044), (June 1985).
- L. Eyberger:
  PCC/SRC Tape Description
  NESC Note 88-83, (June 30, 1988).

11. MACHINE REQUIREMENTS:  1093 Kbytes of memory are required for execution.

The peak working set size to execute the sample problem on a DEC VAX 6000-510 computer is 1268 Kbytes. Using the DISSPLA Library Version 11.0, the corresponding plots were output:
(1) directly on the screen using a DEC VT330 terminal;
(2) on PostScript format files.

13. OPERATING SYSTEM UNDER WHICH PROGRAM IS EXECUTED:  VMS 4.1, 4.2.

The program was installed under VMS 5.5-2. The graphics output was produced using the DISSPLA Library Version 11.0. This proprietary software is not included in the package.

14. OTHER PROGRAMMING OR OPERATING INFORMATION OR RESTRICTIONS

Users must modify the CALL STRTPL statement in subroutine PLOT appropriately to initialize the DISSPLA library in their local computing environments.

15. NAME AND ESTABLISHMENT OF AUTHORS

           R.L. Iman, M.J. Shortencarier, and
           J.D. Johnson
           Sandia National Laboratories, Alburquerque

File name	File description	Records
NESC1057_01.001	Information file of package PCC-SRC	264
NESC1057_01.002	Source code file of program PCC-SRC	1785
NESC1057_01.003	VMS command file to compile,link and run	58
NESC1057_01.004	Input file, read in unit # 1 (data)	1920
NESC1057_01.005	Input file, read in unit # 4 (control cards)	17
NESC1057_01.006	Sample output file (printout results only)	417