home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Graphics Plus
/
Graphics Plus.iso
/
msdos
/
mapping
/
ozgis80
/
ozgis21.tut
< prev
next >
Wrap
Text File
|
1992-01-04
|
14KB
|
185 lines
QUANTISATION
============
Attributes are usually presented to OzGIS as values which have to
be quantized into a number of classes for display.
A maximum of 10 classes can appear in single variate zone displays and
9 classes (a maximum of 3 per variate) in a bivariate display. A maximum
of 4 classes is available for lines and 4 classes for sites.
The quantisation process is the most important aid for the analyst in
understanding the features of the attribute data. The quantisation method
and parameters should be chosen logically according to the purpose of
analysing the data.
The aim is to display the map that best shows the spatial features and
distribution of the data.
Note that the best maps usually have a small number of classes; manipulate the
map to show the data according to requirements. This contrasts with the
production of atlases, where large numbers of colours are used as the purpose
to which the map will be put is not known.
There are other options to change the list of zones to which quantisation is
applied and to change the range of values over which the method operates.
Quantisation Methods
--------------------
The following methods for quantization are available for determining
the class intervals:
(a) Equivalence Classes: numbers are assigned to the attribute values
(possibly with integer round-off). The attribute values should lie
in the range of the maximum number of classes permitted but they
will be scaled for the selected number of classes.
This method enables the quantisation to be carried out on another
system and the resulting class numbers entered instead of attribute
values. A common use is for mapping discrete data e.g. political
parties on election maps.
(b) Quantiles: intervals are computed by assigning the same number of
zones to each class.
This method has often been used to generate choropleth maps, e.g.
deciles. The effect of equal numbers of zones is maps that have
approximate equal areas of each class colour. Such maps are
pretty. Unfortunately quantiles tend to obscure the distribution
of the attribute data.
(c) Equal Value Intervals: intervals are computed from equal
increments over the range of attribute values.
The default quantisation method selected when a map is first
generated is equal value intervals. The advantage of this method
is that the number of zones assigned to each class indicate the
distribution of the data. It is recommended for general purpose
maps and for initial investigations of attribute data.
(d) Refined Equal Value Intervals: intervals are computed from equal
increments over the attribute value range, modified by a
"round-off" procedure (e.g. an increment of 10.12 would become
10.00).
Maps for publications usually have 'nice' values in the legend.
(e) 121 Equal Value Intervals: 121 intervals are computed from equal
increments over the range of attribute values. Only 8 classes are
displayed in the legend, but the colours are assigned over the 121
quantized values to give a "continuous colour" appearance.
This option is only available with standard zone maps.
(f) Interactive Selection of Class Intervals: intervals are selected
by the user by placing crosshairs on a displayed histogram.
(256 colours interactive mode only!)
(g) Mean and Standard Deviation: intervals are determined by dividing
the range of attribute values at the mean value and at specified
offsets from the mean that are multiples of the standard deviation
of the data. The number of classes must be even.
This method has particular application for attribute data from
random populations where the data are expected to have a normal
distribution and hence statistical theorems govern percentages of
population within the classes.
(h) Nested Means: intervals are determined by iterative division of
the range of attribute values at the mean value of the subdivision.
The number of classes must be 2, 4 or 8.
(i) Natural Breaks: intervals are determined by iterative division at
the largest difference between attribute values. The number of
attribute values between differences is user-specified. Hence
class intervals occur at "jumps" in the data.
(j) Specification of Class Intervals: interval values (for a specified
number of classes) are typed in by the user.
Hence data within certain value ranges can be isolated. Suitable
class intervals for hard-copy maps can be selected.
(k) Specification of Numbers Per Class: intervals are determined by
user-specification of the number of zones or sites in each class.
The numbers need only be given for some of the classes; the
remaining zones or sites will be distributed over the remaining
classes during each quantization.
An analyst can isolate data at the extremes of the attribute
distribution by using this method.
(l) Class Number Percentiles: intervals are determined from
user-specified values giving the percentages of the number of zones
within each class.
(m) Class Range Percentiles: intervals are determined from
user-specified values giving the percentage of the total range of
attribute values in each class.
(n) Current Class Intervals: the intervals (and number of classes) are
used to quantize subsequent attributes.
Hence a series of maps can be produced with the same legend which
enables attributes to be compared.
(o) Current Numbers: the number of zones or sites per class (and
number of classes) are used to determine the intervals for
subsequent attributes.
Quantisation Ranges
-------------------
The range of values over which the quantization is applied can be
restricted in all methods. The following options are available for
limiting the range:
- the extremes of all values (default)
- user-specified limits (the user enters the low and high values)
- refined values (i.e., automatically rounded to "nice" values)
- limits fixed at current values for subsequent quantisations
Zones with values outside these limits are assigned the "excluded zone"
value and colour, lines and sites are not displayed.
For example a standard legend for percentage data with value ranges
0,25,50,75 and 100 could be generated by choosing extremes to be 0 and 100
and fixing them, and by using 4 equal value classes.
Quantisation Lists
------------------
Each of the attribute processing streams has an associated list
that holds the names of the items being quantised i.e. zones or lines or
sites. There is one list for a single stream, one zones list for bivariate
maps, and for two streams there is a list of zones and a list of lines or
sites.
Each list selects the items that are to be quantised from the
corresponding attribute file. When a map is generated the lists are set to
all the names if the attribute files (common names in the case of bivariate
maps).
Zone lists can be reset to:
- all zones in current attribute file (single variate)
- all zones common to two attribute files (bivariate)
- the displayed zones
- zones in a names file
Zone lists can also be modified by adding or deleting zone names by
typing in a name or selecting the zone with the cursor (256 colour mode)
Site lists and line lists can be modified by giving the names.
Hence the quantisation can take place for a set of items that is
independent of the displayed, zone lines and sites (although it is
illogical for none to be the same). It is common for the quantisation to
be carried out over a larger geographic area than that being displayed.
Sometimes zones are removed because the attribute data are doubtful e.g.
Census districts with a low population.
Changing attribute files does not change the items whose values are
quantized.