Featherstone & Morgan: Validation of the AUSGeoid98 model in WA 
Figure 5. Histogram of the differences between astronomic-Helmert deviations and AUSGeoid98-derived deviations in the north-south 
component (435 points). Units in arc-seconds. 
bin (arc-seconds) 
Figure 6. Histogram of the differences between astronomic-Helmert deviations and AUSGeoid98-derived deviations in the east-west 
component (435 points). Units in arc-seconds. 
sigma test assuming a normal distribution of the 
deviation differences. Figures 5 and 6 show histograms 
of the differences (including outliers), which are near- 
normally distributed, thus justifying the use of the three- 
sigma test for outlier rejection. If one deviation 
component was determined as an outlier, then both 
components were rejected. However, around fO of the 
outliers were in both deviation components. 
Discussion 
The results in Tables la and lb largely mirror those in 
Featherstone (2006); after the removal of outliers (based 
on the three-sigma criterion), the precision of 
AUSGeoid98-derived vertical deviations in Western 
Australia is roughly one arc-second. Indeed, this is 
commensurate with the estimated precision of the 
astronomically determined deviations. From the error 
analysis in Featherstone & Riieger (2000), this is adequate 
for the reduction and post-processing of current 
terrestrial-geodetic survey observations. As such. 
Table la 
Descriptive statistics of the differences between astronomic- 
Helmert deviations and AUSGeoid98-derived deviations (435 
points, including 15 outliers). Units in arc-seconds 
North-south (^) 
east-west (rj) 
Maximum 
16.9 
9.1 
Minimum 
-7.8 
-10.9 
Moan 
-0.3 
-0.3 
STD 
1.5 
1.5 
Table lb 
Descriptive statistics of the differences between astronomic- 
Helmert deviations and AUSGeoid98-derived deviations (420 
points, excluding 15 outliers). Units in arc-seconds 
North-south (^l 
east-west (q) 
Maximum 
3.0 
3.3 
Minimum 
-4.7 
-4.6 
Mean 
-0.2 
-0.3 
STD 
0.9 
1.1 
147 
