A reliable physics lab report begins with careful data analysis, not with a polished conclusion. This physics data analysis tutorial gives you a reusable workflow for checking measurements, reporting uncertainty, choosing graphs, finding best-fit lines, evaluating residuals, and turning results into a clear lab report. It also includes a small dataset, spreadsheet and Python approaches, and checklists for common experiment types.
Overview
Lab analysis connects raw measurements to a physical model. The goal is not to make data look perfectly linear or to force agreement with a predicted value. The goal is to show what was measured, how reliably it was measured, what relationship the data supports, and where limitations remain.
Use this sequence for most experiments:
- Preserve the original measurements and record units.
- Estimate the uncertainty of each measured quantity.
- Calculate derived quantities with appropriate precision.
- Organize the results in a labeled table.
- Choose graph axes that reflect the model you are testing.
- Fit the data using a physically justified method.
- Inspect residuals and compare the result with the prediction.
- Write a conclusion that refers to evidence, uncertainty, and limitations.
Before analyzing a mechanics experiment, review the relevant model rather than selecting equations by memory. The guides to solving physics problems and using physics formulas with units can help you identify the variables and expected relationship.
Significant figures and uncertainty
Significant figures communicate the precision supported by the measurement. Do not report a calculated result to more decimal places than the input data or instrument resolution justifies. Keep extra guard digits during intermediate calculations, then round the final value.
State uncertainty in the same units as the measured quantity, for example, L = 0.500 ± 0.001 m. The uncertainty may come from instrument resolution, repeated measurements, a stated manufacturer tolerance, or an estimate based on the measurement method. Explain which convention you used.
For a sum or difference, a simple conservative rule is to combine absolute uncertainties:
z = x ± y, so Δz ≈ Δx + Δy.
For a product, quotient, or power, fractional uncertainty is often more useful:
z = xy, so Δz/z ≈ Δx/x + Δy/y; and z = xn, so Δz/z ≈ |n|Δx/x.
These are practical introductory rules. Some courses instead require root-sum-square propagation, especially for independent random uncertainties. Follow the method specified by your instructor and state it in the report. For a broader discussion of uncertainty, graphs, and error analysis, see How to analyze physics lab data.
Checklist by scenario
For any experiment
- Write the research question or testable prediction in one sentence.
- List independent, dependent, and controlled variables.
- Keep raw readings unchanged in a separate table or sheet.
- Include units in column headings, not only in surrounding prose.
- Record the instrument resolution or stated measurement uncertainty.
- Repeat measurements when practical and report the spread or mean appropriately.
- Use dimensional analysis to check derived quantities. The dimensional analysis guide provides a reusable check.
For a linear relationship
Suppose a model predicts y = mx + b. Plot the measured independent variable x horizontally and the measured dependent variable y vertically. Label both axes with quantity and unit, such as “Extension, x (m)” and “Force, F (N).” Add uncertainty bars when they are available.
For example, an investigation of a spring might produce the following simplified dataset:
| Extension x (m) | Force F (N) |
|---|---|
| 0.010 | 0.48 |
| 0.020 | 0.97 |
| 0.030 | 1.46 |
| 0.040 | 1.98 |
| 0.050 | 2.47 |
A best-fit line has a slope of force per extension, so its units are N/m. In this context, the slope represents the spring constant if the model is F = kx. Do not infer meaning from a slope until you have written its units and compared them with the equation.
In a spreadsheet, place x in one column and y in the next. Create a scatter chart rather than a category line chart. Add a linear trendline only if a linear model is justified, display the equation, and record the fit statistic if your course uses one. A high fit statistic alone does not prove that the model is physically correct.
For a nonlinear relationship
First graph the variables in their original form. This shows the actual behavior and can reveal a curve, saturation, or threshold. If theory predicts a transformed relationship, linearize it only after identifying the model. Examples include plotting v2 against radius for some circular-motion relationships, or plotting the period squared against length for an ideal pendulum model.
Record the transformation explicitly. A graph of y2 against x is not the same as a graph of y against x; the slope and its uncertainty also have different meanings. Include both plots when the original data help readers interpret the transformation.
For repeated measurements
Calculate the mean only when repeated readings represent the same quantity under the same conditions. Report the variability using the method required by your course, such as a range, standard deviation, or uncertainty in the mean. Look for drift: readings that steadily increase or decrease may indicate warming, battery discharge, friction changes, or a setup problem rather than random scatter.
Spreadsheet and Python workflow
A spreadsheet is suitable for sorting data, calculating means, propagating simple uncertainties, and making quick graphs. Keep formulas visible and use separate columns for raw values, derived values, and uncertainty calculations. Never overwrite raw readings with rounded or corrected numbers.
Python is useful when there are many measurements or when you need a repeatable analysis. A minimal workflow uses a table of values, then plots the data and calculates a fit:
import numpy as np
import matplotlib.pyplot as plt
x = np.array([0.010, 0.020, 0.030, 0.040, 0.050])
y = np.array([0.48, 0.97, 1.46, 1.98, 2.47])
slope, intercept = np.polyfit(x, y, 1)
fit = slope*x + intercept
plt.scatter(x, y, label="measurements")
plt.plot(x, fit, label=f"fit: y={slope:.2f}x+{intercept:.2f}")
plt.xlabel("Extension (m)")
plt.ylabel("Force (N)")
plt.legend()
plt.show()Check the code by comparing one calculated value with a hand calculation. Software performs arithmetic; it does not decide whether the variables, model, units, or uncertainty method are appropriate.
What to double-check
Tables
- Does every column have a meaningful heading and unit?
- Are raw and processed values clearly separated?
- Are decimal places consistent with the measurement precision?
- Can another reader reproduce one sample calculation?
Graphs and best-fit lines
- Is the graph a scatter plot for measured pairs?
- Are the axes labeled with symbols, quantities, and units?
- Does the scale use the available plotting area without hiding relevant variation?
- Are uncertainty bars included where they affect interpretation?
- Is the best-fit line based on all relevant points rather than just the first and last?
- Does the intercept have a physical interpretation, or should a zero-intercept model be justified separately?
Do not automatically force a line through the origin. A theory may predict zero intercept, but an offset can reveal a zeroing error, background signal, or an incomplete model. Compare the fitted intercept with its uncertainty or with the resolution of the experiment before deciding whether it is significant.
Residuals
A residual is the measured value minus the value predicted by the fit: residual = ymeasured − yfit. Plot residuals against the independent variable when possible. Random residuals scattered around zero support the chosen model more convincingly than a high fit statistic alone. A curve, systematic run of positive and negative values, or increasing spread suggests a model problem, an unaccounted variable, or nonconstant uncertainty.
Comparison with a predicted value
Compare experimental and theoretical values using uncertainty, not just percentage difference. A useful discussion asks whether the difference is larger than the combined uncertainty and identifies plausible reasons for any mismatch. Possible limitations include calibration, parallax, reaction time, friction, heat loss, alignment, finite sensor resolution, or too few data points. Name the limitation that matters for your apparatus rather than listing every possible source of error.
Common mistakes
- Rounding too early: Keep guard digits during calculations and round the reported result at the end.
- Using a line graph for paired measurements: A scatter plot preserves the numerical spacing of the independent variable.
- Confusing precision with accuracy: Repeated values can be tightly grouped while still being offset by calibration error.
- Calling every discrepancy “human error”: Explain the mechanism, such as timing delay or instrument zero offset.
- Ignoring units in slopes: The slope’s units often identify the physical constant being measured.
- Using a best-fit line as proof: A fit summarizes a model; it does not establish causation or remove systematic error.
- Reporting only a percentage error: Include the measured value, uncertainty, reference value, method of comparison, and interpretation.
- Deleting outliers without justification: Investigate the original record first. Remove a point only when there is a documented procedural reason.
- Writing the conclusion before checking residuals: Let the completed analysis determine whether the evidence supports the prediction.
When to revisit
Use this workflow before every new lab, even when the experiment appears familiar. Revisit the uncertainty method when the instrument changes, when measurements are repeated under different conditions, or when the expected relationship includes a transformation. Recheck spreadsheet formulas and Python code whenever columns, units, or calibration values change.
Before submitting, complete this final checklist:
- Confirm that the research question, model, and graph variables agree.
- Check every unit, exponent, and conversion.
- Verify one result by hand.
- Inspect the graph and residuals before interpreting the fit.
- Round values and uncertainties consistently.
- Explain the slope, intercept, and uncertainty in physical terms.
- State whether the data support the prediction and why.
- Name one specific limitation and one realistic improvement.
For related preparation, consult the physics formulas reference and the guide to free-body diagrams when your analysis depends on force models. Save a clean copy of your raw data, calculations, graphs, and code with the final report. That record makes the next experiment faster to analyze and easier to audit.