Gravhögarna på Långön : ett metodtest av statistisk prediktionsmodellering i Norrlands inland

• Main Author: Granholm, Tim
• Format: Bachelor Thesis
• Language: Swedish
• Published: Uppsala universitet, Arkeologi 2020
• Subjects: Prediktionsmodellering; R; Spatial Modellering; Långön; Archaeology; Arkeologi; Northern Sweden
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-414806

In an attempt to improve archaeological predictive modeling, two predictive models were developed through QGIS, Excel, GeoDa and R. These models were then tested with statistical quality tests. The first model was a linear regression model similar to that used by the default predictive models used in GIS software. The second model was a custom exponential model built through R. The two models were compared using MAE and the exponential model yielded slightly improved results. Various problems and opportunities regarding statistics in archaeological work were discovered, and discussed based on this papers findings.The concept of using the econometric methods of spatial predictive modeling was explored and discussed, although deemed unfit for this paper. Although the spatial model was never developed, it was deemed innessecary considering the success of the other models, in particular the exponential model. A few areas could with relative statistical significance be pointed out as likely former settlements, in particular an area to the immidiate west of Långön. The high probability area contains a smaller area previously reported by a civilian to contain rock fragments. Although that report is uncertain, in combination with the model results it is deemed the most promising area discovered during this project.