Paper 5 Planning, Analysis and Evaluation

A plan begins with a testable prediction, named independent and dependent variables and the key variables to standardise.

Link the prediction to a biological mechanism and make the expected direction or pattern explicit.

If increasing light is predicted to increase photosynthetic rate until saturation, the prediction identifies both the trend and its limit.

A vague aim such as “investigate photosynthesis” is not a testable prediction.

A method must vary the independent variable deliberately, measure the response precisely, control key variables, include controls and collect enough results.

Choose volumes, concentrations, intervals and replicates that make the comparison fair; use serial dilution when appropriate.

Prepare a concentration series by proportional dilution, randomise treatment order if drift is possible and measure the same endpoint each time.

A control is a comparison condition, not simply a repeat with no label.

Quality concerns spread and consistency; validity asks whether the design really tests the hypothesis; risk combines hazard severity with likelihood.

Use anomalies, standard deviation, standard error or confidence intervals where appropriate and state precautions.

A narrow spread does not prove validity if the dependent variable is a poor measure of the biological process.

Repeatability and accuracy are related but not identical, and a low-risk procedure can still be invalid.

Present the key values in a clear table or graph, put the independent variable on x and the dependent variable on y, and choose calculations that answer the question.

Use means, percentages, percentage change or rates only when the underlying units and comparison are clear.

A percentage increase should state the baseline used; a rate should state the interval over which it was calculated.

A graph can clarify data but cannot repair an unsuitable scale, missing units or selective omission.

Statistical choice depends on whether data are categoric, ordinal or continuous, on distribution and on the comparison or relationship being tested.

State the null hypothesis, justify the test and interpret probability alongside biological context.

Use chi-squared for observed versus expected categories, a t-test for two means under suitable assumptions, and correlation tests for relationships.

A familiar test is not automatically valid; independence, sample size and distribution matter.

A strong conclusion combines the pattern in raw data, processed values, graphs and statistical results, then states how far the hypothesis is supported.

Explain the biology behind the pattern and make a prediction that follows from the evidence.

If a treatment raises the mean but confidence intervals overlap widely, the conclusion should remain cautious rather than absolute.

Statistical significance alone does not establish the mechanism or practical importance.

Evaluation asks whether anomalies, range, intervals, measurement method, controls and replication allow a confident answer.

Propose changes that directly reduce the identified limitation and explain how confidence would improve.

If the range stops before a plateau, extend it; if intervals are too wide, add intermediate values; if measurement is subjective, use a calibrated method.

Listing every possible limitation weakens an evaluation; prioritise the limitations that affect the conclusion.