Hi all,
I tried a promising workaround for the 'matrix stuff'. However,
there is still a major obstacle:
data ... --odbc isn't allowed within a loop.
Up until now for one period I use loops to generate the SQL
statements and have a bunch of data statements right after the
loop.
This will shorten tremendously if the change in question is
applied.
Is is possible to allow that?
Would be great :-)
Cheers
Leon
Am 28.07.2012 16:04, schrieb Pindar:
Hi there,
the ODBC import tool is a very powerful tool and I appreciate it
much!
However, while working with it I came across some situations when
the option of importing as matrix would be nice
e.g. via data .... --odbc --matrix.
Here are my thoughts:
1) When working with multiple databases (with different
structures) in order to combine information the necessary steps
are: work with 'nulldata' command, creating an appropriate
observation index, use the odbc import, change the series to a
matrix, use 'nulldata --preserve' and start with the next odbc
import. This works fine. The point is, that loops for subsequent
sql statements over the whole script are impossible e.g. when
creating a panel.
2) If the observation string is cryptic like 1/1/FK/2012Q1
importing as matrix would provide an easy way of using the
obs-format option to specify the rownames of the matrix and get
without the need of a 'manual intervention' the 'help information
for the obs-string' into GRETL ( Matrix_Index_1stEntry London 1
...). The usual series name would be the colname.
3) For the right sizing of the index it is crucial to now the
length of an data imported via odbc. If this length is
'endogenous' a count sql statement provides the needed
information. Again if I want to work the script automatically I
need to import a whole series with only one value and have to
rescue this value as scalar (or matrix).
What do you think? Is it possible to add such an option without
much effort? Or should it better be an update to the 'mread'
function?
Cheers
Leon