Theory of Errors, Weights and Most Probable Value - Memory Anchors
Memory techniques transform dry formulas and abstract concepts into vivid, unforgettable mental images. For the PRC Geodetic Engineer board exam, where you must recall dozens of formulas under time pressure, a well-built memory anchor is worth hours of re-reading. This collection uses mnemonics, analogies, micro-stories, and visual chains tailored to Filipino culture and the specific demands of Adjustment Computations. Work through each anchor actively — say it aloud, visualize it, and test yourself. The goal is simple: make every key concept UNFORGETTABLE.
Memory anchors
Three Types of Errors: Blunders, Systematic, Random
Memory aid
Remember the acronym BSR — 'Bawal, Solve, Random.' Blunders are BAWAL (forbidden — eliminate them immediately like a surveyor crossing out a misread rod value). Systematic errors you SOLVE (model and correct, like applying temperature correction to a steel tape). Random errors are just RANDOM — small, unpredictable, and handled by adjustment. Think of BSR as your three-step survey error checklist every time you open a field book.
Why it works
The Filipino word 'Bawal' (forbidden/not allowed) creates an immediate emotional response — blunders must be removed without mercy. The three-step sequence maps perfectly to the action required for each error type, reinforcing both identification and the correct response.
Example usage
Board question asks: 'Which type of error is addressed by least squares adjustment?' Trigger BSR → R = Random → answer is Random errors.
Recall trigger
BSR — Bawal, Solve, Random
Blunders must be eliminated before adjustment
Memory aid
Think of adjustment computations like cooking sinigang. You cannot make the best sinigang if you accidentally add motor oil instead of patis — you must throw out the ruined batch (eliminate the blunder) and start over. No amount of careful simmering (least squares) will fix motor oil in your sinigang. Similarly, a gross misreading of 152.3 m instead of 125.3 m cannot be 'adjusted away' — it must be identified and discarded before any computation begins.
Why it works
Food analogies are universally relatable to Filipinos. The irreversible nature of motor oil in food perfectly mirrors the irreversible contamination a blunder causes in an adjustment dataset.
Example usage
When facing a dataset with a clearly outlying observation, recall the sinigang analogy — remove the blunder first, THEN apply least squares.
Recall trigger
Motor oil in sinigang — you cannot fix it, throw it out.
Systematic errors are modeled and corrected, not averaged out
Memory aid
A systematic error is like a timbangan (weighing scale) in a palengke that always reads 50 grams too heavy. Every single measurement is wrong in the SAME direction. The vendor cannot fix this by weighing things multiple times — each reading is still 50 g too heavy. The fix is to KNOW the error and subtract it. In surveying, thermal expansion of a steel tape always elongates it in the same direction — apply the correction formula, do not rely on averaging.
Why it works
The palengke timbangan is a familiar Filipino everyday object. The consistent directional bias is perfectly mirrored by the scale's constant overreading, making the concept of 'same-direction, correctable' errors intuitive.
Example usage
Exam asks to classify a temperature correction on a steel tape — recall the timbangan analogy → same-direction error → systematic → model and correct.
Recall trigger
Palengke timbangan that always reads 50g too heavy.
Random errors follow the normal distribution — small, sign-varying, most probable value is the mean
Memory aid
Imagine a sharpshooter at a ROTC rifle range. Even the best shot cannot hit exactly the same spot twice. Some shots land a hair left, some right, some slightly high, some low. The errors are small (near the bullseye), random in direction (plus or minus), and if you mark all 100 shots and find the center of the cluster, THAT center is your most probable value. The tighter the cluster, the smaller the standard deviation. This is exactly how repeated surveying observations behave — and the center of the cluster IS the arithmetic mean.
Why it works
The rifle range is a vivid, visceral image. ROTC is familiar to Filipino engineering students. The physical scatter of bullet holes around a bullseye directly mirrors the statistical scatter of random errors around the true value.
Example usage
When asked why the arithmetic mean is the MPV for equal observations, recall the bullet cluster — the center of symmetric scatter is the best estimate.
Recall trigger
ROTC sharpshooter — the center of the bullet cluster.
Most Probable Value (MPV) for equal observations = arithmetic mean
Memory aid
When observations are EQUAL in might, the MEAN is the value that's right. Sum them all, divide by n — that gives the Most Probable Value, amen! Formula: x-bar equals sum of x, divided by n — no guess, no flex.
Why it works
Rhymes encode information in prosodic memory (rhythm and rhyme), a separate cognitive channel from semantic memory. The playful 'amen' ending makes it feel like a conclusion, reinforcing that the mean is the final, definitive answer for equal observations.
Example usage
Five equal-weight measurements given — recall the rhyme → MPV = arithmetic mean → sum all five, divide by 5.
Recall trigger
Equal in might, the MEAN is right.
Residual: v_i = x_i − x̄ (observation minus mean)
Memory aid
A residual is the 'utang' (debt) or 'sobra' (excess) of each observation compared to the best estimate. If the mean is 100.04 m and your measurement is 100.06 m, you 'overpaid' by 0.02 m — that is your residual. If your measurement is 100.02 m, you 'underpaid' by 0.02 m — negative residual. The total of all these utang and sobra, when properly squared and summed, tells you the overall spread of your measurements.
Why it works
Utang (debt) and sobra (excess) are deeply intuitive financial concepts for Filipino students. The positive/negative sign of residuals maps perfectly to overpayment vs. underpayment.
Example usage
Given x=100.06 and mean=100.04, compute residual → recall utang/sobra → v = 100.06 − 100.04 = +0.02 m.
Recall trigger
Utang or sobra versus the mean.
Standard deviation formula: σ = √(Σv²/(n−1))
Memory aid
Remember the phrase: 'Square the Residuals, Sum them, Shrink by One, then Root.' SSSR — Square, Sum, Shrink (divide by n−1), Root. The 'Shrink by One' step is critical — we use n−1 (not n) because we LOST one degree of freedom when we computed the mean from the same data. Think: we used one piece of information to find the mean, so only n−1 pieces remain 'free' to measure spread.
Why it works
The SSSR sequence gives a step-by-step action chain that traces through the entire formula mechanically. 'Shrink by One' is a distinctive phrase that flags the easy-to-forget n−1 in the denominator.
Example usage
Computing σ from residuals: Step 1 square each v, Step 2 sum them, Step 3 divide by (n−1), Step 4 take the square root. SSSR.
Recall trigger
SSSR — Square, Sum, Shrink by One, Root.
Standard deviation of the mean: σ_x̄ = σ/√n
Memory aid
Think of buying pandesal from a bakery. One pandesal has variable weight — sometimes heavier, sometimes lighter. But if you buy a DOZEN and weigh them together, then divide by 12, the average weight is much more consistent (less variable). The more pandesal you average, the more stable your estimate. The improvement is NOT proportional to n — it is proportional to √n. Doubling your sample from 1 to 4 cuts the uncertainty in HALF (√4 = 2), not to one-fourth.
Why it works
Pandesal is a quintessential Filipino daily object. The √n relationship is counterintuitive (students often think more samples = proportionally better), so tying it to the tangible act of buying and averaging bread weights makes the non-linearity concrete.
Example usage
If σ of one reading = 0.015 m and n=5, then σ_mean = 0.015/√5 = 0.0067 m. Recall pandesal analogy to remember it is √n in the denominator.
Recall trigger
More pandesal averages → uncertainty shrinks by √n, not n.
Weight is inversely proportional to variance: w = 1/σ²
Memory aid
A veteran geodetic engineer, Sir Ben, and a nervous first-year utility man both measure the same baseline. Sir Ben's σ = 0.01 m; the utility man's σ = 0.05 m. In the adjustment meeting, Sir Ben's measurement gets weight w = 1/0.0001 = 10,000 while the utility man's gets w = 1/0.0025 = 400. Sir Ben's result is trusted 25 times more! This is the key insight: the more precise you are (smaller σ), the LOWER your variance, and the HIGHER your weight. Weight = trust = 1/σ². The SQUARE is critical — never use 1/σ alone.
Why it works
The senior engineer vs. junior worker dynamic is immediately relatable in a Philippine field survey context. The numerical comparison (25× more trust) makes the impact of the σ² relationship viscerally clear. The warning 'never use 1/σ alone' directly addresses the most common board exam pitfall.
Example usage
Given σ₁=0.02 and σ₂=0.01, compute w₁=1/0.0004=2500, w₂=1/0.0001=10000. Always square σ first.
Recall trigger
Sir Ben vs. utility man — trust is inversely proportional to σ² (not σ).
Weight for leveling is inversely proportional to route distance: w ∝ 1/K
Memory aid
Imagine two jeepney routes between Manila and Bulacan. Route A is 2 km of smooth, direct road; Route B is 8 km of winding, potholed road. Which route accumulates more bumps (errors)? The long route. For differential leveling, every kilometer adds more random error — so the longer the route, the LESS reliable the result. Weight is INVERSELY proportional to distance: shorter route = more weight = more trust. w ∝ 1/K where K is in kilometers.
Why it works
The jeepney analogy grounds an abstract statistical relationship in familiar Filipino transportation. The 'more bumps on a longer road' intuition perfectly mirrors error accumulation in leveling.
Example usage
Routes of 2, 1, 4 km → w₁=0.5, w₂=1.0, w₃=0.25. Shortest route has highest weight, just like the shortest jeepney route has fewest potholes.
Recall trigger
Short jeepney route = fewer bumps = higher weight. w = 1/K.
Weighted mean formula: x̄_w = Σ(w_i × x_i) / Σw_i
Memory aid
A weighted mean is like a class deliberation where professors vote on a failing student's final grade. Professor A (weight 3) says 75, Professor B (weight 1) says 65. The grade is not simply (75+65)/2 = 70 — Professor A counts three times as much! So: (3×75 + 1×65)/(3+1) = (225+65)/4 = 72.5. The student gets 72.5, not 70. In surveying, a more precise instrument IS Professor A — it counts more in the final answer. Always multiply each value by its weight, sum them, then divide by the SUM OF WEIGHTS.
Why it works
The grading committee scenario is immediately recognizable to Filipino students. The consequence (different final grade) makes the math meaningful. The parallel to instrument precision reinforces the geodetic application.
Example usage
w₁=2500, x₁=100.02; w₂=10000, x₂=100.05 → x̄_w = (2500×100.02 + 10000×100.05)/(2500+10000) = 100.044 m.
Recall trigger
Professor with more authority (weight) controls the grade — multiply, sum, divide by total weight.
Probable error = 0.6745σ
Memory aid
Remember: 'Six-Seven-Four-Five times sigma gives you probable error.' Or memorize 0.6745 as 'sixty-seven forty-five' — like a phone number. The probable error is the radius of the interval that has exactly 50% probability of containing the true value. Think of it as the 50-50 zone. And the magic number 0.6745 is simply the z-value where the cumulative normal distribution reaches 75% (giving a 50% symmetric interval).
Why it works
Chunking 0.6745 into '67-45' (like a 4-digit PIN or phone extension) leverages numerical chunking memory. The '50-50 zone' label provides a conceptual hook that explains WHAT the probable error means.
Example usage
If σ = 0.020 m, then PE = 0.6745 × 0.020 = 0.01349 m ≈ 0.013 m.
Recall trigger
PIN 6745 × σ = probable error (the 50-50 zone).
Weight is also directly proportional to number of observations: w ∝ n
Memory aid
Think of a barangay survey. If Kagawad A measured a distance once and Kagawad B measured it ten times, who do you trust more for the average? Kagawad B — because ten measurements, when averaged, give a much more reliable result. More measurements = higher weight. This is a direct (not inverse) relationship: w ∝ n. Contrast this with leveling distance where w ∝ 1/K — there it is INVERSE. The key difference: more observations = more trust (direct); more leveling distance = less trust (inverse).
Why it works
The barangay kagawad setting is a local government context familiar to Filipino students. The direct comparison of once vs. ten measurements makes the proportionality concrete. The explicit contrast with the leveling rule prevents confusion.
Example usage
Observation set A: measured 4 times; Set B: measured 1 time. Relative weights: w_A = 4, w_B = 1.
Recall trigger
Kagawad who measured 10 times is trusted more: w ∝ n (direct).
Why n−1 (not n) in the standard deviation denominator
Memory aid
Imagine you are in a barkada of 5 friends planning to split a bill. One friend (the mean) is chosen to 'hold' the total. Now only 4 friends (n−1) have 'freedom' to decide how much they contribute freely — the last person's contribution is completely determined by what the others paid. In statistics, once you compute the mean FROM your data, you have 'used up' one degree of freedom. Only n−1 observations are truly FREE to vary. This is why σ uses n−1: you are estimating from a SAMPLE, and one freedom was consumed making the mean.
Why it works
The barkada bill-splitting scenario is deeply relatable to Filipino college students. The concept of 'one person whose amount is fixed by the others' perfectly illustrates a degree of freedom in a non-technical way.
Example usage
Computing σ from 6 observations: denominator is 6−1 = 5, not 6. Recall the barkada — the last person has no choice.
Recall trigger
Barkada bill: the last person's share is not free — n−1 degrees of freedom.
Precision vs. Accuracy in the context of errors
Memory aid
Picture two dartboards. Dartboard 1 (PRECISE but not accurate): all darts clustered tightly together, but far from the bullseye — like a systematic error (consistent but wrong direction). Dartboard 2 (ACCURATE but not precise): darts scattered around the bullseye with no pattern — random errors, but the center is correct. Dartboard 3 (PRECISE AND ACCURATE): darts tightly clustered ON the bullseye — this is the goal of adjustment. Systematic errors destroy ACCURACY (shift the cluster away from bullseye). Random errors destroy PRECISION (scatter the cluster). Adjustment addresses RANDOM errors.
Why it works
Dartboard diagrams are the classic visual for precision/accuracy. The link to error types (systematic = shifts cluster, random = scatters cluster) creates a complete conceptual mapping that is both visual and analytical.
Example usage
When a question asks which type of error causes measurements to be consistently above the true value — recall dartboard 1 — that is systematic error.
Recall trigger
Three dartboards: tight-but-off (systematic), scattered-but-centered (random), tight-and-on (goal).
The weighted mean is pulled toward the observation with the highest weight
Memory aid
Think of a tug-of-war between two teams. Team 1 (w=2500) pulls toward 100.02 m. Team 2 (w=10000) pulls toward 100.05 m. Team 2 is FOUR TIMES STRONGER. The rope (the weighted mean) gets pulled much closer to Team 2's side — resulting in 100.044 m, which is much nearer to 100.05 than to 100.02. The weighted mean always 'leans' toward the more precise (higher-weight) observation.
Why it works
Tug-of-war (piko ng lubid) is a familiar Filipino school and fiesta game. The physical intuition of a stronger team pulling the rope toward their side perfectly captures how higher weights dominate the weighted mean.
Example usage
Verify a weighted mean calculation: if one weight is much larger, the result should be close to that observation. If it is not, recheck your computation.
Recall trigger
Tug-of-war: the stronger team (higher w) wins — the mean leans toward them.
Redundancy (number of observations minus number of unknowns) determines degrees of freedom
Memory aid
Remember: r = n − u, where r = redundancy, n = number of observations, u = number of unknowns. Mnemonic: 'RENU' — Redundancy Equals N minus U. Redundancy is what makes adjustment possible — without it, you have an exact solution and no ability to detect errors. In leveling, measuring a loop gives redundancy; in a triangle, measuring all three angles gives 1 redundancy (r=3−2=1, since the third angle is determined by the first two).
Why it works
RENU sounds like a Filipino nickname, making it easy to remember. The triangle example grounds the abstract definition in a concrete geometric case familiar from surveying courses.
Example usage
Four angle observations in a triangle → n=4 (3 angles + 1 check), u=2 (only 2 are independent) → r=4−2=2 degrees of freedom.
Recall trigger
RENU: Redundancy = n − u.
The mean minimizes the sum of squared residuals (least squares principle)
Memory aid
Engineer Maria is adjusting elevation data. She tries three different 'best estimates': 45.10, 45.13, and 45.17 m. Each time, she computes Σv². She notices that Σv² is smallest when she uses the arithmetic mean (45.13 m). She tries shifting slightly up or down — Σv² always increases. The arithmetic mean is the UNIQUE value that makes Σv² an absolute minimum. This is NOT a coincidence — it is the mathematical definition of least squares, and it is WHY the mean IS the least-squares solution for direct, equal-weight observations.
Why it works
The micro-story of Engineer Maria shows the least-squares principle as an empirical discovery rather than a handed-down formula. The act of 'trying different values' and observing Σv² grow makes the minimization principle intuitive.
Example usage
When a board question asks 'what property does the arithmetic mean satisfy?' — recall Maria's experiment → the mean minimizes Σv².
Recall trigger
Engineer Maria trying different estimates — the mean always wins the Σv² competition.
Common pitfall: using 1/σ instead of 1/σ² for weight
Memory aid
Remember the warning phrase: 'WEIGHT has POWER TWO — never one will do!' If σ = 0.02, the weight is 1/(0.02)² = 2500, NOT 1/0.02 = 50. Using 1/σ (power one) gives a completely wrong answer and is the NUMBER ONE board exam trap in this topic. The square comes from the statistical derivation of the normal distribution's exponent. Tattoo it in your mind: w = 1/σ², always σ-SQUARED.
Why it works
The rhyming warning ('POWER TWO — never one will do') is emphatic and targets the specific common error. Explicitly naming it the 'NUMBER ONE board exam trap' raises its threat level and ensures students pay special attention.
Example usage
Board exam: σ₁=0.03, σ₂=0.02. Wrong: w₁=33.3, w₂=50. Correct: w₁=1111, w₂=2500. Always square σ.
Recall trigger
POWER TWO — never one will do! w = 1/σ².
Improving precision: σ_mean improves as √n, not as n
Memory aid
To halve the error in your mean, take FOUR times the readings you have seen. To cut it to one-third, measure NINE. The square root rules — and that is the line! Formula: σ_mean = σ/√n. So: n=4 gives half the error (√4=2); n=9 gives one-third (√9=3); n=16 gives one-quarter (√16=4). Quadruple the work for half the gain — that is the cruel reality of random error reduction.
Why it works
The rhyme highlights the diminishing returns of increasing sample size, which is a key insight students often miss. The numerical examples (4, 9, 16) provide instant recall of the non-linear relationship.
Example usage
If σ=0.024 m and you take 9 measurements instead of 1, σ_mean = 0.024/√9 = 0.024/3 = 0.008 m — three times more precise.
Recall trigger
To HALVE the error, take FOUR times the readings. √n rules.
Revision game
I am not blunder, not systematic — I scatter small and change sign freely. The normal curve is my home. Who am I?
Answer
Random Error
Memory link
A4 — ROTC sharpshooter scatter pattern; BSR anchor A1 (the R in BSR)
I am the 'best guess' for equally-reliable direct measurements. I minimize the sum of squared residuals. What am I?
Answer
Arithmetic Mean (Most Probable Value)
Memory link
A5 rhyme ('Equal in might, the MEAN is right') and A18 Engineer Maria micro-story
I express how much one observation deviates from the best estimate. My formula is simple: reading minus mean. In Filipino, I am either 'utang' or 'sobra'. What am I?
Answer
Residual (v_i = x_i − x̄)
Memory link
A6 — utang/sobra analogy
I represent how much you trust a measurement. I am the reciprocal of variance — never the reciprocal of standard deviation alone. What am I?
Answer
Weight (w = 1/σ²)
Memory link
A9 Sir Ben vs. utility man story; A19 'POWER TWO — never one will do' warning
I am the confidence interval that has exactly 50% probability of containing the true value. My magic number is 0.6745. What am I?
Answer
Probable Error (PE = 0.6745σ)
Memory link
A12 — PIN 6745 mnemonic
For a leveling network, I assign higher trust to shorter routes. I am inversely proportional to route length K. What relationship describes me?
Answer
Leveling Weight Rule: w ∝ 1/K
Memory link
A10 — jeepney route analogy (shorter route = fewer potholes = higher weight)
I tell you the precision of the average, not of a single reading. I shrink as you take more measurements, but only as fast as the square root of n. What am I?
Answer
Standard Deviation of the Mean (σ_x̄ = σ/√n)
Memory link
A8 pandesal analogy and A20 rhyme ('To halve the error, take FOUR times the readings')
I explain why the denominator in the single-observation standard deviation is n−1, not n. I am the number of independent pieces of information remaining after the mean is computed. What concept am I?
Answer
Degrees of Freedom (n−1)
Memory link
A14 — barkada bill-splitting story (last person's share is fixed by the others)
Formula mnemonics
x̄ = Σx / n
Mnemonic
SUM it all, COUNT them, DIVIDE — the arithmetic MEAN will be your guide.
When to use
When all observations have equal reliability (equal weights). This gives the Most Probable Value for direct, equally-weighted measurements.
What each part means
x̄ = most probable value (mean); Σx = sum of all n observations; n = total number of observations
v_i = x_i − x̄
Mnemonic
RESIDUAL = READING minus the MEAN. 'R = R minus M.' Each reading's utang or sobra from the mean.
When to use
After computing the mean, to find how far each observation deviates. Used as input to compute standard deviation.
What each part means
v_i = residual of the i-th observation; x_i = individual observation value; x̄ = computed arithmetic mean (MPV)
σ = √(Σv² / (n−1))
Mnemonic
SSSR: Square the residuals, Sum them, Shrink by one (n−1), Root the result.
When to use
To measure the spread/precision of a single observation in a set of n repeated direct measurements. Use n−1 (sample standard deviation).
What each part means
σ = standard deviation of a single observation; v = individual residuals; n−1 = degrees of freedom (n observations minus 1 consumed by the mean)
σ_x̄ = σ / √n
Mnemonic
Mean is MORE precise: sigma over ROOT-n. More readings = smaller sigma-bar (but improvement slows — it is square root, not n).
When to use
To express the precision of the mean (MPV) rather than a single reading. Always smaller than σ by factor 1/√n.
What each part means
σ_x̄ = standard deviation of the mean; σ = standard deviation of a single observation; √n = square root of the number of observations
PE = 0.6745 σ
Mnemonic
PIN 6745 × sigma = Probable Error (the 50-50 boundary zone).
When to use
When a problem asks for the probable error of a single observation or of the mean. For the mean: PE_mean = 0.6745 × σ_x̄.
What each part means
PE = probable error (50% confidence interval half-width); 0.6745 = the z-score at the 75th percentile of the normal distribution; σ = standard deviation
w_i = 1 / σ_i²
Mnemonic
WEIGHT has POWER TWO — never one will do! w = 1 over sigma-SQUARED.
When to use
When observations have different precisions (different σ values). Assign weights before computing weighted mean.
What each part means
w_i = weight of observation i; σ_i² = variance (square of standard deviation) of observation i. Higher precision (smaller σ) → smaller variance → larger weight.
w ∝ 1/K (leveling)
Mnemonic
Shorter jeepney route = fewer potholes = higher weight. w is INVERSE to distance K.
When to use
Specifically for differential leveling routes to assign relative weights when route lengths differ.
What each part means
w = weight of a leveling observation; K = length of the level route in km. The proportionality constant cancels in the weighted mean calculation.
x̄_w = Σ(w_i × x_i) / Σw_i
Mnemonic
WEIGHTED MEAN: multiply each value by its weight, SUM the products, then DIVIDE by the SUM OF WEIGHTS (not the count n).
When to use
When observations have different reliabilities (different weights). Replaces simple mean. Note: divide by Σw_i, NOT by n.
What each part means
x̄_w = weighted mean (best estimate with unequal weights); w_i = weight of observation i; x_i = value of observation i; Σw_i = sum of all weights
Quick recall chains
Steps to Compute MPV from Direct Observations
Recall test
Without looking, list the 6 steps to compute the MPV and precision of direct observations in order.
Memory chain
Think of a surveyor named 'Ben Adds Standard Residuals, Sigma Sigma': B=Blunders out, A=Apply corrections, S=Sum and divide (mean), R=Residuals computed, S=Standard deviation, S=Standard deviation of mean. Or use the story: BEN'S BOSS ARRIVES — B=Blunders, B=Bad systematics corrected, A=Arithmetic mean, R=Residuals, S=Standard deviation, S=Sigma of mean.
Items to remember
- Step 1: Identify and eliminate blunders
- Step 2: Apply systematic error corrections
- Step 3: Compute arithmetic mean (MPV = Σx/n)
- Step 4: Compute residuals (v_i = x_i − x̄)
- Step 5: Compute standard deviation σ = √(Σv²/(n−1))
- Step 6: Compute standard deviation of the mean σ_x̄ = σ/√n
Three Types of Errors and Their Remedies
Recall test
Name the three error types and the correct action for each, using only the BSR trigger.
Memory chain
BSR: Bawal (eliminate), Solve (correct), Randomize (adjust). Remember: each error type demands a DIFFERENT action — no single cure works for all three.
Items to remember
- Blunders → Eliminate (remove from dataset)
- Systematic → Model and Correct (apply formulas)
- Random → Adjust (least squares, weighted mean)
Rules for Assigning Weights
Recall test
State the weight formula for: (a) observations with different standard deviations, (b) leveling routes of different lengths, (c) groups with different numbers of repeated measurements.
Memory chain
The WEIGHT RULES: 'Vary Square, K Inverse, Numbers Direct.' Think: σ-Squared gives inverse weight; K (km) gives inverse weight; n (count) gives direct weight. Three rules, two inverses, one direct.
Items to remember
- w = 1/σ² (inversely proportional to variance)
- w ∝ 1/K for leveling (inversely proportional to route distance)
- w ∝ n for repeated observations (directly proportional to count)
- w ∝ 1/PE² (inversely proportional to square of probable error)
Key Numbers to Memorize for Errors
Recall test
Without looking: What multiplier of σ gives the probable error? What probability corresponds to 1σ? What denominator is used in sample standard deviation?
Memory chain
Phone number chain: 6745 for Probable Error, 6827 for one-sigma (68.27%), 9599 for 95% (remember as '95 is almost there'), 99 for 99%. And always n-MINUS-ONE for the denominator.
Items to remember
- 0.6745σ = Probable Error (50% probability)
- σ = Standard Deviation (68.27% probability for 1σ interval)
- 1.9599σ = 95% confidence interval (2σ ≈ 95%)
- 2.5758σ = 99% confidence interval
- n−1 = degrees of freedom for sample standard deviation
Weighted Mean Computation Steps
Recall test
Describe all 5 steps to compute a weighted mean from memory.
Memory chain
DWMSD: Determine weights, Weight-multiply, Make the sum of products, Sum the weights, Divide. Say it as 'Diwata Makes Sexy Deals' — Determine, Multiply, Sum products, Sum weights, Divide.
Items to remember
- Step 1: Determine or compute each weight w_i
- Step 2: Multiply each observation by its weight: w_i × x_i
- Step 3: Sum the products: Σ(w_i × x_i)
- Step 4: Sum the weights: Σw_i
- Step 5: Divide: x̄_w = Σ(w_i × x_i) / Σw_i
Diagrams
Theory of Errors, Weights and MPV — Complete Concept Map
A complete hierarchical mind map showing all major concepts in the chapter and their relationships. Use this as a study overview and to check that no concept has been missed.
Decision Flowchart: Which Formula to Use for Observations
A step-by-step decision flowchart guiding students through the correct formula selection process. Follow this flowchart for any board exam problem involving repeated measurements. BSR order is embedded: Blunders first, Systematics second, then Random error adjustment.
Weight Assignment: Which Rule Applies?
A decision flowchart for choosing the correct weight formula based on what information is provided in the problem. Includes warnings about the most common pitfall (using 1/σ instead of 1/σ²).
Historical Development of Error Theory in Surveying
A timeline showing the historical development of the theory of errors and least squares, from Gauss to modern Philippine geodetic practice under PRS92. Helps students appreciate that these formulas have a 200-year history of geodetic application.
Distribution of Board Exam Questions by Topic in This Chapter
An estimated distribution of board exam question types for this chapter based on past PRC Geodetic Engineer exam patterns. Weighted mean computation and standard deviation together make up 55% of questions — prioritize these topics.