IB Physics Internal Assessment: Data Processing

The Internal Assessment (IA) is your opportunity to demonstrate the skills of a practicing physicist. While a compelling research question is essential, your ability to collect, process, and present data with rigorous scientific integrity is what separates a good IA from a great one. Mastering data processing is not just about following steps; it’s about constructing a logical, evidence-based argument that directly addresses your question and showcases your understanding of measurement and uncertainty—a core pillar of the IB Physics syllabus.

From Raw Measurements to Processed Data

Your investigation begins with raw data: the direct, unaltered measurements you take from your experiment. These must be recorded immediately in a clear, well-designed table. Each column should have a descriptive heading, the relevant SI unit, and an indication of the measurement uncertainty. For example, if you use a digital stopwatch that reads to hundredths of a second, your uncertainty might be $\pm 0.01\text{ s}$. For analog instruments like a ruler, a common rule is to take half of the smallest division (e.g., $\pm 0.5\text{ mm}$ for a 1 mm scale). Presenting raw data this way immediately shows the examiner you understand the limitations of your apparatus.

Once recorded, you often need to process this raw data into a form suitable for analysis. This involves calculations. For instance, if you measure the period $T$ of a pendulum ten times, you would calculate the mean period $\bar{T}$. More complex processing might involve calculating derived quantities, like using $T=2\pi \sqrt{\frac{l}{g}}$ to find an experimental value for $g$. Every calculated value must be accompanied by its propagated uncertainty. The rules for this are fundamental:

• For addition/subtraction: absolute uncertainties add. If $C=A+B$, then $\Delta C=\Delta A+\Delta B$.
• For multiplication/division: percentage uncertainties add. If $C=A\times B$, then $\frac{\Delta C}{C}=\frac{\Delta A}{A}+\frac{\Delta B}{B}$.
• For powers: multiply the percentage uncertainty by the power. If $C=A^n$, then $\frac{\Delta C}{C}=|n|\frac{\Delta A}{A}$.

A processed data table should include all calculated quantities with their final absolute and percentage uncertainties, clearly demonstrating this error propagation.

Graphical Analysis and Linearization

A graph is not merely a presentation tool; it is a powerful analytical instrument. Your goal is to plot processed data in a way that allows you to test the theoretical relationship suggested by your research question. Often, this requires linearization. If theory suggests a power law like $y=k{x}^n$, plotting $y$ against $x$ will yield a curve. Instead, you can take logarithms of both sides:

$$\log (y)=\log (k)+n\log (x)$$

Plotting $\log (y)$ against $\log (x)$ should then produce a straight line with gradient $n$ and y-intercept $\log (k)$. This transforms a complex analysis into a simpler one of finding a line of best fit.

Every data point on your graph must include error bars representing its uncertainty. These are crucial for two reasons. First, they visually represent the precision of your data. Second, they allow you to draw a line of best fit and, just as importantly, worst acceptable lines (lines of maximum and minimum reasonable slope that still pass through all error bars). The uncertainty in the gradient (slope) can be estimated as half the difference between the maximum and minimum slopes: $\Delta m=\frac{|m_{\text{max}}-m_{\text{min}}|}{2}$. This graphical method of determining uncertainty is highly valued in the IA.

Deriving Conclusions and Evaluating the Investigation

Your conclusion must be a direct, quantified answer to your research question, derived from the graphical analysis. State your final result (e.g., "the experimental value for gravitational acceleration is $9.76\pm 0.15{\text{ m s}}^{-2}$") and compare it to the accepted literature value. A percentage discrepancy calculation is useful here:

$$\text{Percentage Discrepancy}=\frac{|\text{Experimental Value}-\text{Accepted Value}|}{\text{Accepted Value}}\times 100\%$$

Crucially, compare this discrepancy to your total percentage uncertainty. If the discrepancy is less than or similar to your uncertainty, you can claim agreement within experimental limits. If it is significantly larger, you must acknowledge a systematic error.

The evaluation section is where you demonstrate critical thinking. Systematic errors shift all measurements in one direction (e.g., a slightly mis-calibrated sensor, consistent parallax error). Discuss their likely direction and impact on your results. Random errors are seen in the scatter of points around your line of best fit and are reflected in your error bars. You must also suggest specific, realistic improvements. Vague statements like "use better equipment" are insufficient. Instead, propose a concrete change ("use a light gate connected to a data logger instead of a manual stopwatch to reduce human reaction time error") and explain how it would reduce a specific uncertainty or systematic error you identified, and how it would lead to a more precise or accurate result.

Common Pitfalls

1. Uncertainties as an Afterthought: Simply stating "human error" or listing uncertainties in a separate section without propagating them through calculations and onto graphs is a major loss of marks. Uncertainty analysis must be integrated into every stage of your processing.
2. Ignoring the Research Question in Analysis: It is common to see students perform linearization and gradient calculations correctly but then fail to link the gradient's physical meaning back to their original question. Always state what physical quantity your gradient (or y-intercept) represents.
3. Poor Graphical Practice: Drawing a line of best fit that connects the dots, failing to include error bars, or not using graph paper (or software equivalent) with appropriate scaling (where data occupies more than half of each axis) will undermine an otherwise good analysis. The graph is a core piece of evidence.
4. Generic, Non-Physics Improvements: Suggesting improvements like "be more careful" or "repeat the experiment more times" does not score well. Improvements must be technically relevant. For example, to reduce air resistance effects, propose a vacuum pump; to improve temperature measurement, suggest a thermocouple with a data logger instead of a mercury thermometer.

Summary

• Record raw data meticulously in tables with clear units and justified measurement uncertainties for every quantity.
• Process data systematically, performing all calculations with proper error propagation to generate final processed values with their associated uncertainties.
• Use graphs as analytical tools, employing linearization techniques where necessary, and always plotting data with error bars to determine a best-fit line and estimate the uncertainty in your gradient.
• Derive a quantitative conclusion from your graph, comparing your result to an accepted value using percentage discrepancy and contextualizing it within your calculated percentage uncertainty.
• Evaluate your method critically, distinguishing between systematic and random errors, and proposing specific, actionable improvements that target identified weaknesses in your experimental design or technique.