Lecture 6

Question 1

Uses of GIS

Answer 1

Lots of choropleth maps are produced to show new data
Consider what is shown, scale (population), colour scheme
Red and green should never be seen (red and green colour blindness)

Question 2

Aims

Answer 2

Error and uncertainty in data and models 
Sources of error and uncertainty 
Managing error and uncertainty
Understanding MAUP
- Scale
- Zones

Question 3

Error- why it matters

Answer 3

GIS analysis relies on quality data- but also sensible technicians who can identify inaccuracies and the source of error.
Error can render the analysis useless, meaningless or cause inappropriate decisions to be made based on faulty outputs.
We bring together lots of diverse datasets (a huge benefit), which create more chances for error to be introduced.

Question 4

Introduction to error and uncertainty

What do we mean by ‘error’?

Answer 4

Everyday language:
A mistake
Quantitative sciences:
Difference between a predicted or measured value and the true value
Error = measured value - true value
Error = predicted value - true value

Question 5

Limiting Digitizing Error
Heads-up digitizing (using a mouse) can lead to user error
Potential sources of digitizing error:

Answer 5

Overshoots and undershoots
Dangling segment
Sliver polygons

Question 6

Closeness of output to reality is unknown…

Answer 6

Therefore errors introduce uncertainty into outputs (maps, statistics etc…)

Question 7

Accuracy

Answer 7

The extent to which an estimated data value represents its true value

Question 8

Precision

Answer 8

The dispersion of a measured or predicted value also specificity (decimal places)

Question 9

Bias

Answer 9

Is any systematic difference between the true value and the measured or predicted value (e.g overprediction)

Question 10

Three key types of accuracy in GIS: Positional, attribute and topological

Answer 10

Positional accuracy: are features in the right place?
Attribute accuracy: do features have the correct value (continuous/ categorical)?
Topological accuracy: Do features relate to each other

Question 11

Common errors

Answer 11

Misalignment errors in overlay
Aggregation/ disaggregation
Classification (especially of continuous data to categorical)
Errors will propagate through

Question 12

Understanding scale effects

Answer 12

Fotheringham and Wong (1991) explain the increase in the coefficient of correlation through the fact that aggregation into larger and larger area is equivalent to smoothing and therefore a reduction in variance for the variables of interest.
This outcome is also affected by the presence of spatial dependence among the initial observations: positive spatial autocorrelation moderates the fall in variance from smoothing, while negative spatial autocorrelation exacerbates it.

Question 13

Zonation

Answer 13

How you draw the boundary of an area may influence the resulting measure of health outcomes
Often zones are created for convenience, based on historic boundaries
Major differences in correlations

Question 14

Sources of error and uncertainty in GIS analysis

Answer 14

Data collection
Data encoding
Data editing/ conversion
Data analysis