Weighted Observations: Using Statistical Reliability to Find the MPV

Overview

In high-precision surveying, not all measurements are created equal. Some observations are taken by more experienced observers, some use superior instrumentation, and some are averages of larger data sets. To combine these varied measurements into a single result, surveyors use Weighted Observations to find the Most Probable Value (MPV) 1, 2.

Why This Matters

If you treat a measurement with a standard error of $\pm 5''$ the same as one with $\pm 1''$, you degrade the quality of your final result. Weighting allows the surveyor to mathematically favor more reliable data, ensuring that the most precise observations have the greatest influence on the final coordinates or elevations 2, 3.

Theory

The weight of an observation ($W$) is a measure of its relative reliability compared to other observations in the same set. By definition, weight is inversely proportional to the variance ($\sigma^2$) of the measurement 3, 4.

The Most Probable Value (MPV)

The MPV is the closest approximation to the true value that can be achieved from a set of data. For a series of weighted observations $L_1, L_2, \dots, L_n$ with corresponding weights $W_1, W_2, \dots, W_n$, the MPV is the Weighted Mean 1, 2: $\text{MPV} = \frac{\sum (L \times W)}{\sum W}$

Mathematical Principles

1. Relationship Between Weight and Error

The fundamental rule for assigning weights based on standard error ($\sigma$) is 4: $W \propto \frac{1}{\sigma^2}$ This means that if measurement B is twice as precise as measurement A (half the error), it receives four times the weight 3.

2. Standard Error of the Weighted Mean ($S_{\bar{x}_w}$)

The precision of the calculated weighted mean is given by 4: $S_{\bar{x}_w} = \pm \sqrt{\frac{\sum (W \times r^2)}{(n-1) \sum W}}$(Where $r$ is the residual of each observation from the weighted mean).

Field Workflow

Step-by-Step Example

Problem: An angle is measured by three observers. Determine the MPV based on their recorded means and standard errors 2:

• Observer A: $89^\circ 54' 36'' \pm 0.7''$
• Observer B: $89^\circ 54' 42'' \pm 1.2''$
• Observer C: $89^\circ 54' 33'' \pm 1.0''$

1. Calculate Weights ($1/S^2$):
   • $W_A = 1/0.7^2 = 2.04$
   • $W_B = 1/1.2^2 = 0.69$
   • $W_C = 1/1.0^2 = 1.00$
2. Simplify Ratios (optional, but good for checks):
   • Total weight $\sum W = 3.73$
3. Multiply Seconds by Weights:
   • $36'' \times 2.04 = 73.44''$
   • $42'' \times 0.69 = 28.98''$
   • $33'' \times 1.00 = 33.00''$
4. Sum and Divide:
   • $\sum (LW) = 135.42''$
   • $\text{MPV Seconds} = 135.42 / 3.73 = 36.3''$
5. Result: MPV = $89^\circ 54' 36.3''$ 2.

Practical Tips

• The "Fractional" Shortcut: When calculating the MPV of an angle or large number, only apply the weighting to the differing seconds or decimals to keep the numbers manageable 2.
• Rejection of Outliers: Before weighting, any observation whose residual is greater than a specified rejection criterion (e.g., $3S$ or Chauvenet’s criterion) should be discarded 3, 5.

Common Mistakes

• Weighting by Error ($1/S$): A common beginner error is weighting by the error itself rather than the square of the error ($1/S^2$). Weighting must be based on variance 3, 4.
• Assuming Equal Weight: Never assume observations are equal if the conditions (equipment, weather, distances) changed significantly between sets.

FAQ

Conclusion

Weighting is the tool that transforms a collection of measurements into a unified, defensible result. By linking weight to the inverse of variance, the surveyor ensures that the most precise field work dictates the quality of the final engineering design.

References

Schofield, W. (2001). Engineering Surveying. 5th ed. Butterworth-Heinemann.