Bad Data in Your IB Chemistry EE: What to Do Next

Bad experimental data does not automatically ruin an IB Chemistry Extended Essay. The important question is not whether every point follows a perfect curve. It is whether you can determine what the data supports, explain what it does not support, and evaluate the investigation honestly.

The current IB Extended Essay guide for first assessment 2027 explicitly says that negative results are as valid as positive ones when they are critically evaluated in relation to the research question. It also requires conclusions to be supported by the data. That means you should not manufacture a clean result, delete awkward measurements without justification, or claim that your hypothesis was confirmed when the evidence is inconclusive.

First, Decide What “Bad” Means

Students use “bad data” to describe several different problems. Diagnose yours before changing the analysis.

This distinction matters. A rejected hypothesis is not the same as an invalid experiment. In contrast, a concentration calculated from the wrong dilution factor is a processing error that should be corrected.

A 300% Coefficient of Variation Needs Context

The coefficient of variation (CV), also called relative standard deviation when expressed as a percentage, compares the standard deviation with the mean:

CV = standard deviation ÷ mean × 100%

A CV of 300% may indicate that the spread is very large relative to the mean. But it can also become unstable when the mean is close to zero. The NIST guidance on coefficient of variation notes that CV is intended for ratio-scale data with a meaningful zero and is sensitive to small changes when the mean is near zero.

Before treating a large CV as proof that the experiment failed, ask:

1. Is the variable measured on a ratio scale with a meaningful zero?
2. Is the mean close to zero because positive and negative values cancel?
3. Did you calculate CV from raw readings or from a transformed value where it is less meaningful?
4. Is one unusual replicate inflating both the standard deviation and CV?
5. Do the raw data and error bars tell the same story?

Report the statistic that answers your research question. Do not add CV merely because it looks sophisticated. The broader principles in this Chemistry IA data-analysis guide also apply to experimental EE data: show enough processing for the reader to follow the result, use units consistently, and interpret each statistic rather than listing it.

Audit the Data Before You Rewrite the Essay

Work from the original observations, not a cleaned spreadsheet.

1. Verify transcription and calculations

Check every suspect value against your lab record. Recalculate dilution factors, calibration conversions, gradients, reaction rates, percentage uncertainties, and unit conversions. Keep the original value visible if you correct a genuine transcription or formula error, and record what changed.

2. Plot raw replicates, not only means

A mean can hide useful patterns. Plot individual trials alongside the mean and uncertainty. You may discover that one concentration was consistently unstable, that variance increased with concentration, or that readings drifted over time.

3. Revisit calibration and instrument limits

For spectrophotometry, check whether absorbance values fell within the calibrated range, whether a blank was used consistently, and whether the calibration relationship was appropriate across the full range. For titration, consider endpoint judgment and burette resolution. For rate experiments, consider whether the sampling interval was fast enough to capture the change.

4. Map variation to plausible chemical or procedural causes

Avoid generic statements such as “human error affected the result.” Name a mechanism and connect it to the observed direction or spread. Examples include:

• temperature drift changing the rate constant;
• incomplete mixing producing inconsistent local concentrations;
• volatile solvent loss increasing the effective concentration;
• an interfering species affecting absorbance;
• inconsistent timing creating a larger effect for faster trials;
• glassware tolerance propagating through serial dilutions.

If you are still planning the investigation, the experimental-design section of the Chemistry EE lab-to-paper guide can help you identify these risks before the main data collection.

Do Not Delete an Outlier Just Because It Is Inconvenient

An outlier may be a mistake, random variation, or a scientifically meaningful observation. NIST guidance on outlier detection cautions against automatically deleting an outlying observation and recommends combining graphical inspection with appropriate statistical reasoning.

If you exclude a value, document:

• the pre-existing or scientifically justified rule used;
• the evidence that the measurement was compromised;
• the result with and without the value where useful;
• the effect of exclusion on the conclusion.

“It spoiled the trend” is not a valid reason. With only a few replicates, formal outlier tests can also be fragile, so discuss the decision with your supervisor rather than applying a test mechanically.

Should You Repeat the Experiment?

Repeat data collection when it is safe, permitted, feasible, and likely to address an identified weakness. Do not repeat trials only until the preferred result appears.

A repeat is most defensible when:

• a calibration or preparation mistake invalidated a set;
• the instrument operated outside its useful range;
• the method did not standardize an important control variable;
• there are too few replicates to characterize random variation;
• a pilot reveals a better independent-variable range before the final run.

If the deadline prevents new collection, say so in the evaluation only if it helps explain the evidence available. Your essay can still make a bounded conclusion: the data may show no reliable relationship under the tested conditions, or it may suggest a trend that cannot be distinguished confidently from experimental variation.

How to Write the Results, Discussion, and Evaluation

Keep the three jobs distinct.

Results: show what happened

Present raw and processed data clearly, with headings, units, uncertainties, sample sizes, and readable graphs. The 2027 guide requires standardized presentation and enough methodological detail for replication.

Discussion: explain what the pattern means

Connect the result to chemical theory and relevant published evidence. Address the size and direction of the effect, not just whether a trendline slopes upward. If uncertainty overlaps substantially or the relationship is weak, use cautious language such as “suggests,” “is consistent with,” or “does not provide strong evidence for.”

Evaluation: explain how the method shaped the evidence

Prioritize limitations by their likely effect on the conclusion. For each important limitation, explain:

1. what happened;
2. whether it mainly affected precision, accuracy, or validity;
3. how it may have changed the data;
4. what specific modification would address it.

This is stronger than a long catalogue of generic errors. It also aligns naturally with the analysis and evaluation expected by the Chemistry EE criteria.

Example: Turning a Weak Evaluation Into a Useful One

Weak: “The high standard deviation was caused by human error. More trials should be completed.”

Stronger: “Absorbance varied most at the two highest concentrations. Those samples were above the upper region represented by the calibration standards, so converting absorbance to concentration required extrapolation. This reduces confidence in the calculated values and may explain why their spread increased. A repeat should extend the calibration range, dilute unknowns into that range, and use the same blank before each measurement series.”

The stronger version identifies evidence, explains its consequence, and proposes a modification that directly addresses the limitation.

A Practical Recovery Checklist

• Preserve the original raw data.
• Verify calculations, units, and calibration.
• Plot individual replicates and uncertainty.
• Check whether CV is meaningful for the variable and mean.
• Investigate unusual points without deleting them automatically.
• Compare the observed pattern with chemical theory and sources.
• Limit the conclusion to what the evidence supports.
• Rank the most consequential methodological weaknesses.
• Propose specific, realistic improvements.
• Ask your supervisor before recollecting or excluding data.

Unexpected results are not a loophole around sound chemistry, but neither are they an automatic failure. A strong EE treats uncertainty as part of the evidence and builds an honest argument around what the investigation actually found.

Related Reading and Useful Tools

• IB Chemistry EE: From Lab Work to Final Paper
• Chemistry Extended Essay Criteria Guide
• IB Chemistry IA: Mastering Data Analysis and Presentation
• Get rubric-aligned Chemistry EE feedback