$P = \dfrac{a^3 b^2}{c^{1/2}\, d}$
Percentage error formula:
$$\frac{\Delta P}{P}\% = 3\frac{\Delta a}{a}\% + 2\frac{\Delta b}{b}\% + \frac{1}{2}\frac{\Delta c}{c}\% + \frac{\Delta d}{d}\%$$(Official NEET 2025 answer key: 13%)
For $P = x^a y^b / z^c$: relative error $\dfrac{\Delta P}{P} = a\dfrac{\Delta x}{x} + b\dfrac{\Delta y}{y} + c\dfrac{\Delta z}{z}$. The coefficient of each term = absolute value of the power of that variable. For addition/subtraction $P = x \pm y$: absolute errors add: $\Delta P = \Delta x + \Delta y$. For multiplication/division: relative errors add. For power $P = x^n$: $\Delta P/P = n(\Delta x/x)$. Negative powers (division) also contribute positively to error.
Random error: unpredictable fluctuations. Reduced by averaging many measurements. Mean absolute error $\overline{\Delta a} = \frac{1}{n}\sum|a_i - \bar{a}|$. Systematic error: consistent bias. Same direction every measurement. Examples: zero error, calibration error. Cannot be reduced by averaging. Must be identified and corrected. Gross errors: blunders (misreading instrument). Eliminated by careful repetition. Limiting error: maximum possible error in a single measurement = least count of instrument.
Addition/subtraction: result has same number of decimal places as the least precise operand. Example: $2.5 + 0.032 = 2.5$ (not 2.532). Multiplication/division: result has same number of significant figures as the least precise operand. Example: $2.5 \times 3.14 = 7.9$ (2 sig figs). Rounding: if next digit $\geq 5$, round up; if $< 5$, truncate; if exactly 5 and preceding digit is odd: round up; if even: leave. Significant zeros: $1.00 \times 10^3$ has 3 sig figs; $1000$ has ambiguous sig figs.
Dimensions: $[M]$, $[L]$, $[T]$, $[A]$, $[K]$, $[mol]$, $[cd]$. Every valid physical equation must be dimensionally homogeneous (same dimensions on both sides). Uses: (1) Check equation validity (necessary but not sufficient). (2) Derive relations (limited by unknown dimensionless constants). (3) Convert units between systems. (4) Find dimensions of physical constants. Example: $[G] = M^{-1}L^3T^{-2}$ from $F = GMm/r^2$. Cannot distinguish $\sin\theta$, $\cos\theta$, $e^x$ (all dimensionless).
Metre scale: LC $= 1$ mm. Vernier callipers: LC $= 1$ MSD $-$ 1 VSD (usually 0.1 mm or 0.02 mm). Screw gauge: LC $= $ pitch/circular scale divisions (usually 0.01 mm). Stopwatch (digital): LC $= 0.01$ s. Thermometer: LC $= 0.1°$C (lab). Maximum error per reading $\approx \pm$LC/2 (estimated midpoint). For better precision: use instrument with smaller LC. Accuracy vs precision: accuracy = closeness to true value; precision = reproducibility.
Order of magnitude: power of 10 that represents the quantity. Express number in form $a \times 10^n$ where $1 \leq a < 10$. Order = $n$. Useful for quick checks. Examples: size of atom $\sim 10^{-10}$ m, hydrogen nucleus $\sim 10^{-15}$ m, Earth radius $\sim 10^7$ m, Sun-Earth distance $\sim 10^{11}$ m. Fermi estimation: back-of-envelope calculation using reasonable assumptions. Example: "How many piano tuners in a city?" (famous Fermi problem). Used in NEET for very large/small quantity comparisons.
Vernier zero error: when jaws closed, Vernier 0 not at main scale 0. Positive zero error: Vernier 0 to the right $\Rightarrow$ instrument reads high $\Rightarrow$ subtract. Negative zero error: Vernier 0 to left $\Rightarrow$ reads low $\Rightarrow$ add. Correction: True $=$ Observed $-$ Zero error. Screw gauge: zero error when flat faces touching. Positive: thimble 0 below reference line $\Rightarrow$ reads high $\Rightarrow$ subtract. Negative: thimble 0 above line $\Rightarrow$ reads low $\Rightarrow$ add. Backlash error: loose threads in screw gauge, avoided by always rotating in same direction.
For $n$ measurements $a_1, a_2, ..., a_n$: Mean $\bar{a} = \sum a_i/n$. Mean absolute error $\Delta\bar{a} = \sum|a_i-\bar{a}|/n$. Relative error $= \Delta\bar{a}/\bar{a}$. Standard deviation $\sigma = \sqrt{\sum(a_i-\bar{a})^2/(n-1)}$ (for small sample). Standard error of mean $= \sigma/\sqrt{n}$ (reduces with more measurements). Result expressed as $\bar{a} \pm \Delta\bar{a}$. Confidence interval: 68% of measurements within $\bar{a} \pm \sigma$ for Gaussian distribution. NEET uses simple absolute/relative/percentage error calculation, not advanced statistics.