Error MessagesErrors can arise in a variety of ways.
The SHAZAM output file must be carefully checked for error messages.
To begin, the output messages that follow SHAZAM will continue processing commands after errors are encountered and all subsequent results will be unreliable. The SHAZAM command file must be appropriately modified so that the resulting output file is completely free of error messages. General types of errors are:
Each of these is explained in more detail below. Some errors may be relatively easy to correct while others can be difficult. The help of a SHAZAM advisor may be needed. SHAZAM work should be started well in advance of any deadline so that unanticipated problems, that are the reality of any computing endeavour, can be dealt with in a satisfactory manner.
|
|
The variable name CONSUME
is spelled incorrectly . . .
|
Note that the $ character points to the location of the problem.
The above mistake is corrected, but now the right-hand parenthesis is omitted . . .
|
Problems occur with calculations that involve divide by zero, or the logarithm of a negative number. When these are encountered, SHAZAM sets the result to a missing value code and prints a warning message on the output file. This may be a signal to a problem that needs to be corrected.
An example of some SHAZAM output where an attempt was made to take the logarithm of a negative number follows.
|
Any statistical analysis is limited by the quality of the data set. Sometimes it may not be possible to get a satisfactory solution. For example, the following error message puzzles many users.
|
This means that a matrix inversion was required and SHAZAM found that
the matrix is singular.
With the OLS
command this means that the explanatory variables
have perfect collinearity.
Sometimes results are obtained that do not make much sense. This could occur if an error message appeared earlier on. However, it may also occur even when the output file has no error messages. Some users think that there is a bug in SHAZAM. However, SHAZAM is a well-tested system and so other possibilities need to be carefully considered.
The command file should be carefully reviewed. After READ
commands and GENR
commands it is a good idea to put
a PRINT
command that lists the data. The data listing
should be inspected to make sure that the data set is constructed
correctly. When this is verified then the PRINT
commands
can be removed to make the output file easier to browse through.