SHAZAM reply to McCullough's paper
B. D. McCullough, "Econometric Software Reliability: EViews,
LIMDEP, SHAZAM and TSP", Journal of Applied Econometrics,
Vol. 14, 1999, pp. 191-202.
A number of algorithm improvements have been implemented in
response to the tests proposed in the software review. These
include:
- Section 2 and Section 3.1, pp. 192-4 -- the algorithm for calculating
a sample variance has been improved.
The results for the Michelso dataset are available.
The results for the NumAcc4 dataset are available.
- Section 4, p. 199 -- the random number generator has been modified
so that a "pass" is obtained for all DIEHARD tests. The
DIEHARD results that were produced with
a sample of random numbers generated by SHAZAM can be studied.
- Section 5, p. 200 -- numerical accuracy of critical values
for statistical distributions has been improved.
An example of the 1% critical value of the
F(3,3) distribution can be viewed.
Some documentation of algorithms is provided.
Example
A description of the Michelso dataset
is available. SHAZAM commands for calculating summary statistics
follow.
SAMPLE 1 100
READ (MICHELSO.txt) Y
STAT Y / MEAN=YBAR STDEV=S
FORMAT(F30.16)
PRINT YBAR S / FORMAT
* Get the sample autocorrelation at lag 1
ARIMA Y / NLAG=1 ACF=ACF1
PRINT ACF1 / FORMAT
STOP
|
The SHAZAM output is below.
|_SAMPLE 1 100
|_READ (MICHELSO.txt) Y
UNIT 88 IS NOW ASSIGNED TO: MICHELSO.txt
1 VARIABLES AND 100 OBSERVATIONS STARTING AT OBS 1
|_STAT Y / MEAN=YBAR STDEV=S
NAME N MEAN ST. DEV VARIANCE MINIMUM MAXIMUM
Y 100 299.85 0.79011E-01 0.62427E-02 299.62 300.07
|_FORMAT(F30.16)
|_PRINT YBAR S / FORMAT
YBAR
299.8523999999999000
S
0.0790105478190507
|_* Get the sample autocorrelation at lag 1
|_ARIMA Y / NLAG=1 ACF=ACF1
ARIMA MODEL
NUMBER OF OBSERVATIONS = 100
...NOTE..SAMPLE RANGE SET TO: 1, 100
IDENTIFICATION SECTION - VARIABLE=Y
NUMBER OF AUTOCORRELATIONS = 1
NUMBER OF PARTIAL AUTOCORRELATIONS = 1
0 0 0
SERIES (1-B) (1-B ) Y
NET NUMBER OF OBSERVATIONS = 100
MEAN= 299.85 VARIANCE= 0.62427E-02 STANDARD DEV.= 0.79011E-01
LAGS AUTOCORRELATIONS STD ERR
1 -1 0.54 0.10
LAGS PARTIAL AUTOCORRELATIONS STD ERR
1 -1 0.54 0.10
|_PRINT ACF1 / FORMAT
ACF1
0.5351996686212657
|_STOP
| |
A comparison of the SHAZAM results with the certified values follows.
Inaccurate digits for the SHAZAM results are underlined.
Std.Dev. Autocorrelation at lag 1
SHAZAM 0.0790105478190507 0.5351996686212657
-- ---
certified 0.0790105478190518 0.535199668621283
Univariate Summary Statistics
A description of the NumAcc4 dataset
is available. SHAZAM commands for calculating summary statistics
follow.
SAMPLE 1 1001
READ (NUMACC4.txt) Y
STAT Y / MEAN=YMEAN VAR=YVAR
FORMAT(F20.8)
PRINT YMEAN YVAR / FORMAT
STOP
|
The SHAZAM output is below.
|_SAMPLE 1 1001
|_READ (NUMACC4.txt) Y
UNIT 88 IS NOW ASSIGNED TO: NUMACC4.txt
1 VARIABLES AND 1001 OBSERVATIONS STARTING AT OBS 1
|_STAT Y / MEAN=YMEAN VAR=YVAR
NAME N MEAN ST. DEV VARIANCE MINIMUM MAXIMUM
Y 1001 0.10000E+08 0.10000 0.10000E-01 0.10000E+08 0.10000E+08
|_FORMAT(F20.8)
|_PRINT YMEAN YVAR / FORMAT
YMEAN
10000000.20000010
YVAR
0.01000000
|_STOP
| |
Statistical Distributions
SHAZAM commands for calculating the 1% critical value of the
F(3,3) distribution follow.
SAMPLE 1 1
GEN1 ALPHA=0.01
DISTRIB ALPHA / TYPE=F DF1=3 DF2=3 INVERSE
STOP
|
The SHAZAM output is below.
|_SAMPLE 1 1
|_GEN1 ALPHA=0.01
|_DISTRIB ALPHA / TYPE=F DF1=3 DF2=3 INVERSE
F DISTRIBUTION- DF1= 3.0000 DF2= 3.0000
MEAN= 3.0000 VARIANCE= 0.0000 MODE= 0.20000
PROBABILITY CRITICAL VALUE PDF
ALPHA
ROW 1 0.10000E-01 29.457 0.48920E-03
|_STOP
| |
[SHAZAM Guide home]
|