22 ^
ratios were computed for each test and compared with the tabulated value of
the F-statistic at the 90Z confidence interval. These variance ratios are
given in Table 8. Note that no significant difference exists between cal-
culated and tabulated values of F for Test Cases 1A, 1B and 1C indicating
that a valid distribution of random normal deviates were applied to the plate
coordinates. On the other hand, comparison of calculated with tabular values
of F for Test Cases 2A, 2B, and 2C, indicates the presence of a significant
amount of systematic error in the plate coordinates at the 90% confidence
interval.
2. Supplementary Tests
Participants 1, 7 and 8 performed adjustments using twenty-five points
per photograph. Results from this series of tests are given in Table 9.
Special features to note in connection with these tests are: (a) Partici-
pant 1 utilized Control Array A' illustrated in Figure 9; (b) Participant 7
made 4 adjustments with 60% sidelap; and (c) Participant 7 employed 3 and
9 units per strip in the adjustment of the block.
3. Comparisons
A comparison of the average RMSE's in discrepancies of Participants 1, 2,
3 and 7 (Sequential) with Participants 8, 9, and 10 (Simultaneous) is pos-
sible by examining Table 10. Also given in this table are the percent
changes in average RMSE's. Note that use of simultaneous solutions results
in a decrease in position and elevation errors, of from 26 to 68% and from
20 to 51%, respectively.
On the average, Participants 2 (Sequential) and 8 (Simultaneous )
achieved the lowest discrepancies in their respective groups. A comparison.
of RMSE's in discrepancies for these two participants is.given in Table 11.
The percent change through use of the direct simultaneous Solution (Partici-
pant 8) is from 43 to -48 in position and from -9 to -57 in elevation. Note,
that although Participant 2 did use a sequential procedure, the adjustment
was by a simultaneous linear transformation of all models in the block.
This procedure is a relatively new development in sequential methods.
The percent changes in discrepancies due to an increase in control
point density are listed in Table 12(a) and (c) while Table (b) shows per-
cent increase in control points.
The effects of the simulated residual systematic perturbations are
illustrated in Table 13 where the increases and percent changes in posi-
tion and elevation are listed for each participant. Correlation of results
for a given participant's test case in Table 13 with the appropriate
Tables 1A, 1B, 1C or 2A, 2B, 2C enables determination of the degree
equation used in the adjustment. For example, Participant 4 shows a Amy
of +4.96 meters from Test 1C (random only) to 2C (random + sys.). Ref-
erence to Tables 1C and 2C reveals that in Test 1C a 2nd degree adjustment
in Z was performed while in Test 2C a 3rd degree adjustment was made in Z.
Obviously use of the higher degree equation to compensate for discrepancies
caused by systematic errors is well advised.
