Chemistry · Lesson

Systematic effects and uncertainty in classical methods

Trace volume, mass, endpoint and reference effects through a quantitative model. Distinguish a supplied known error from uncertainty and use the correct net denominator.

Back to the Chemistry map

By the end, you can

Propagate independent reading uncertainty through a difference.
Calculate uncertainty of a stipulated titrimetric concentration ratio.
Use net mass rather than gross mass as a relative-uncertainty denominator.
Compare a fixed endpoint shift across different equivalence volumes.

The measurement function identifies the relevant effects

A classical result is calculated from quantities such as reference value, titrant concentration, volume difference, specimen mass and stoichiometric factor. Endpoint deviation, recovery and composition can require corrections or additional uncertainty terms. Identify the actual function before deciding which effects cancel or which denominator belongs to a relative value.

An error is a difference from a relevant true or reference value. Measurement uncertainty characterizes the values attributed using available information; it is not the actual unknown error. A known significant effect can be corrected, while uncertainty of the correction and unresolved significant effects still need appropriate treatment. Repeating the same procedure does not make a shared effect vanish.

A perfectly common additive offset can cancel in a difference

In an invented volume-reading model, initial indication is 1.20 mL and final indication 21.20 mL, giving difference 20.00 mL. Give each reading independent standard uncertainty 0.0300 mL. Standard uncertainty of the difference is √(0.0300²+0.0300²)≈0.0424264 mL, or 0.212132% relative to the difference.

Separately stipulate the same additive +0.100 mL zero offset in both readings. It cancels exactly: (21.20+0.100)−(1.20+0.100)=20.00 mL. This algebraic cancellation depends on the effect being identical and additive in both indications.

A scale-factor error, position-dependent calibration effect or endpoint shift is a different model and does not automatically cancel. Nor should two independent reading contributions be treated as the same common offset. Covariance belongs to the actual relationship among effects.

A ratio retains uncertainty in concentration and both volumes

For an optional deeper independent-input toy, choose a specified 1:1 reaction with C_A=c_tV_t/V_s. Given c_t=0.100 mol L⁻¹, V_t=20.0 mL and V_s=25.0 mL, the analyte concentration is 0.0800 mol L⁻¹. The units in the volume ratio cancel.

Assign standard uncertainties u(c_t)=0.000200 mol L⁻¹, u(V_t)=0.0500 mL and u(V_s)=0.0300 mL. Their relative standard uncertainties are 0.00200, 0.00250, 0.00120. Under a small-error first-order model with independent inputs and an exact stipulated stoichiometric factor, relative variances add.

Relative standard uncertainty is √(0.00200²+0.00250²+0.00120²)≈0.00341906420. Thus u(C_A)≈0.000273525136 mol L⁻¹. This specified result omits other effects such as endpoint-model uncertainty or a shared standard; those would need their own assessment. A two-proton factor or correlated volumes would change the appropriate model.

Relative uncertainty belongs to the quantity actually used

In a separate weighing toy, gross mass is 30.100 g and tare is 30.000 g. Net mass is 0.100 g. With independent standard uncertainties 0.0000300 g for each weighing, u(net)=√2×0.0000300≈0.0000424264 g.

Relative standard uncertainty of net mass is approximately 0.0424264%. Dividing that same uncertainty by the gross 30.100 g would instead give approximately 0.000140952%, a misleadingly small number for an analyte calculation based on the net mass. The subtraction has made the relevant denominator much smaller.

A perfectly shared additive zero offset can cancel in this difference, but independent weighing effects and other mass-dependent contributions must retain their actual relationships. The provided values are a calculation model, not a balance specification or an operating procedure.

The same absolute endpoint deviation has a different relative effect

Suppose an invented known positive endpoint-minus-equivalence shift is +0.0800 mL, with constant characterized titrant concentration and the same specified stoichiometric factor. If endpoint volume is used without correction, the inferred analyte amount shift relative to the true equivalence inventory is ΔV/V_eq.

For equivalence volumes 5.00, 10.0, 20.0, 40.0 mL, relative shifts are +1.60%,+0.800%,+0.400%,+0.200%. The same absolute shift is a larger fraction of a smaller inventory. This supplied-known-error puzzle is not a measured endpoint bias or a universal reason to change specimen amount.

An actual endpoint correction needs evidence of sign, magnitude, stability and relevant uncertainty. Indicator, reaction equilibrium and instrument response can produce different relationships; a correction from one matrix or regime cannot silently be applied everywhere.

A fixed endpoint shift changes relative amount differently

Equivalence volume / mL	Endpoint shift / mL	Relative amount shift
5.00	+0.0800	+1.60%
10.0	+0.0800	+0.800%
20.0	+0.0800	+0.400%
40.0	+0.0800	+0.200%

Original known-error toy with endpoint volume exceeding equivalence by 0.0800 mL, constant titrant concentration and a fixed specified stoichiometric factor. Positive shifts follow ΔV/V_eq when the uncorrected endpoint is used. No actual method bias, specimen recommendation or accepted tolerance is assigned.

Report the model rather than a precision claim alone

Retain the reference assignments, readings, corrections, endpoint rule, weighed identity, stoichiometry and uncertainty evaluation. State whether uncertainty is standard or expanded and identify any coverage factor and its justified interpretation.

The result’s scope follows those documented choices. A small repeatability SD, canceled common offset or favorable relative number does not establish the absence of other systematic effects or adequate uncertainty for the intended use.

Worked examples

Separate independent noise from a common offset

Use 1.20/21.20 mL indications and independent standard uncertainties 0.0300 mL each.

Difference is 20.00 mL.
Standard uncertainty is approximately 0.0424264 mL.
Relative standard uncertainty is approximately 0.212132%.
A perfectly shared additive +0.100 mL offset cancels algebraically.

Cancellation of one common effect does not remove independent effects.

Combine the declared relative variances

Use the stipulated 1:1 concentration ratio and independent uncertainties.

Concentration is 0.0800 mol L⁻¹.
Relative standard uncertainty is approximately 0.00341906420.
Standard uncertainty is approximately 0.000273525136 mol L⁻¹.

A first-order uncertainty calculation under the declared input scope.

Divide by the net mass

Use gross 30.100 g, tare 30.000 g and independent u=0.0000300 g per reading.

Net mass is 0.100 g.
u(net)≈0.0000424264 g.
Relative u(net)≈0.0424264%.
Using gross mass as denominator would give approximately 0.000140952%.

The relevant denominator is the net quantity used in the result.

Normalize the same shift to different inventories

Use known +0.0800 mL shift with the four equivalence volumes.

At 5.00 mL, shift fraction is 0.0800/5.00=1.60%.
At 10.0, 20.0, 40.0 mL, shifts are 0.800%, 0.400%, 0.200%.
The inferred amount shift is positive in this chosen model.

A fixed absolute deviation has an inventory-dependent relative effect.

Try it yourself

Use the same known +0.0800 mL endpoint shift at equivalence volume 8.00 mL. Find the relative amount shift.
Show the answer and reasoning
+1.00%.
0.0800/8.00=0.0100; the sign follows the chosen positive volume shift.
In a new difference model, both independent reading standard uncertainties are 0.0200 mL and the delivered difference is 10.0 mL. Find standard and relative standard uncertainty.
Show the answer and reasoning
Approximately 0.0282843 mL; 0.282843%.
The independent variance sum gives √2×0.0200, then divide by the actual difference.
Does the cancellation of a common additive zero offset prove that a common multiplicative scale error also cancels?
Show the answer and reasoning
No.
A factor applied to both readings multiplies their difference; it is a different measurement model.
Is standard uncertainty the numerical value of the actual unknown error?
Show the answer and reasoning
No.
It characterizes attributed-value dispersion using available information; the actual error is a separate quantity.

What to carry forward

Start with the actual measurement function.
Keep known error and uncertainty distinct.
Common effects cancel only under the right model.
Relative uncertainty uses the result’s actual denominator.
Endpoint corrections need evidence and scope.

References

Teaching resources and reference material supporting this lesson’s explanations and examples.