3. Discussion

3.1 Resolving the problems

As is evident from the above discussion, the main problem with this whole process was that it was quite time intensive. A further, technical, problem concerned the transfer of the digitised artefact data into GIS in a quick and easy manner. Several options were tested during the initial stages of the project, but there appears to be two main procedures for adding data to a GIS project. The first is by a direct connection to the quantitative summary data with associated XY coordinates in Access, using the OLE DB provider for ODBC driver in the ArcCatalog component of GIS itself, and the second is by exporting the Access select queries containing the XY coordinates and quantitative data as separate Dbase IV format tables and then importing the latter into GIS.

Direct connection to the Access data through ArcCatalog is potentially a useful means by which to add tabular data to ArcGIS, but, in the early stages of the project, we discovered that much of the data could not be plotted, especially in the pie-charts which were the most useful chart style in ArcGIS for this type of data. This was because Access generated null values in the queries' data rows where no occurrences of a specified category were found. ArcGIS 8.3, used for these earlier stages, ignored any row, and thereby any provenance location that contained null values in its pie-chart plots. This effectively excluded most of the data in many of the queries. One of the potential solutions to this problem was to use the Nz function in Access to replace the null values with zeros. This is not a standard SQL command, however, and so it could not be read by ArcGIS.

Thus, the insertion of null values into Access queries, and these values' lack of SQL consistency, caused problems when trying to plot these artefact data directly in ArcGIS. The solution to both these problems was either to use Dbase IV files, exported from Access and imported separately to ArcGIS with the null values replaced by zeros manually, or to use the ArcCatalog component of ArcGIS to create a direct link to Access, which performed the operation automatically. While exporting the DBase IV files from Access and then adding them as layers in ArcGIS meant the production of a large number Dbase IV files, it provided a robust and relatively simple solution to this technical problem.

This null-value translation problem was encountered only during the analysis of Vetera I (See Allison et al. 2004, section 7.2.3). We used ArcGIS 9 for the other sites and found that this newer version could resolve some of these difficulties. In effect, rows containing null values were no longer ignored by this version, and the manual insertion of zeros was no longer required.

However, another complication in a project that uses interpretative data is that these data often needed to be re-evaluated and re-interpreted. A need for the re-interpretation and re-classification of the data, and therefore the need for changes to the main catalogue, was often only evident after the data had been plotted and analysed in GIS. For example, subsequent studies (e.g. Böhme-Schönberger 2002) have required changes to some of the categories (e.g. the gender categorisation of some brooches), and hence changes to the artefact distribution plots in earlier publications, as indicated in Allison in this volume (section 5.1). Also, after a series of queries on the data from Oberstimm had been run and plotted, it was apparent that the building phases, to which the artefacts were assigned, needed to be refined and the sub-phases included. In these two cases all the queries relating to the gender category or to the phasing of artefacts had to be re-constructed and re-exported into Dbase for importing into GIS. GIS often did not directly recognise the additional or changed information, in these new queries, so the data had to be re-formatted and re-imported. This process could have been simplified by changing the data in the Dbase files but it was important to maintain an original up-to-date and complete set of data that could be relied upon in any future analyses. This was held in the formatted artefact spreadsheet, in Excel, which is an easy form in which to adjust any data, and which contains all the data from the original printed catalogues, not just those being queried. This is important if the data tables are to be useful in future analyses which seek to answer other research questions.

In summary, although null values are an issue, the direct linking of Access to ArcGIS is not advisable. The process of data re-evaluation means ongoing correction to the data and subsequent changes to the SQL codes used in the catalogue. The ongoing processes of renewed large-scale querying and re-analysis appear to be the only valid methodology for dealing with these types of interpretative data.

PREVIOUS NEXT CONTENTS HOME