Data Quality - continued! Standards and Positional Accuracy

Continuing with data quality and accuracy assessments, this week we used the Positional Accuracy Handbook for the National Standard for Spatial Data Accuracy (NSSDA) to perform an assessment and write a formal statement using their guidelines. From our assigned reading it states that there are five important criteria when considering data quality, which may have an effect on decision making, liability concerns, and efficiency of productivity: position accuracy, attribute accuracy, logical consistency, completeness and lineage. 

During our second lab, we completed an assessment of two data sets by referencing the difference between where they intersected, and where the intersection should take place based on high resolution imagery of the area. Below are the resulting 20 city test points:

The study area includes the city of Albuquerque, and it is easy to notice quickly that the city's centerlines (green) appear to be much more accurate than ESRI's street polylines (purple), when using imagery for comparison. ESRI's street-lines appear warped and are often shifted northwest from the roadways. It was recommended to find intersections at 90 degree angles for the best result, which was not difficult to do. Once the points were created, coordinate data was added to the shapefiles and exported using Table to Excel. It was then copied and placed into a spreadsheet created based off the NSSDA accuracy assessment guidelines, allowing calculation of the RMSE and NSSDA values. 

The following is the formal statement of the results, showing that ESRI's sample is indeed less accurate than the city of Albuquerque's samples:

• Albuquerque City Streets Sample: 
  
  
• Positional Accuracy: Tested 227.7 feet horizontal accuracy at 95% confidence level
  
  
• ESRI's TeleAtlas Streets Sample: 
  
  
• Positional Accuracy: Tested 365.3 feet horizontal accuracy at 95% confidence level

This assignment was good practice in useful statistical analysis that can ensure your data is of the best quality possible. But we are not done! One more week focused on data quality, coming soon...