Appendix 2 6, 7, 8; 9GE0/04 - Drawing conclusions and evaluating an enquiry
Learn how to answer an enquiry question with a supported conclusion, evaluate the evidence and methods, and judge what findings mean beyond the sampled setting. All examples and checks are original formative practice, not conclusions for your own assessed investigation.
Answering the question with evidence
A conclusion answers the investigation's question by drawing together the relevant findings rather than repeating each result in isolation.
Start with what the question actually asks: difference, relationship, explanation or success against a criterion. Answer its sub-questions, then combine them to reach a judgement about the main question. Cite the decisive numerical patterns and qualitative findings, including evidence that weakens the preferred explanation. Use the theory to interpret results, not to replace them. Do not introduce new unsupported data at the conclusion stage.
Consider an original illustrative enquiry, invented for this lesson: “Is perceived access to a hypothetical riverside space the same for daytime and evening users?” In a convenience sample, 18 of 24 daytime respondents and 8 of 20 evening respondents described access as easy. Three invented evening comments mentioned infrequent buses; two mentioned a difficult crossing. No comparison of the same individuals was made, and respondents were asked on only one weekday.
The sample proportions are 75% and 40%, a difference of 35 percentage points. A defensible practice conclusion is that access was perceived more positively in the daytime sample, with evening transport and crossing concerns suggesting reasons for the contrast. It cannot claim that every resident experiences this difference or that bus frequency alone caused it. The groups may differ in age, mobility, journey purpose or familiarity; these characteristics were not controlled. The question concerns perceived access, not objectively measured journey time.
Synthesis
Combining relevant findings and ideas to support a coherent answer, while explaining how they relate and where they conflict.
“The hypothesis is proved” is usually too strong for a bounded observational enquiry. State that evidence supports, partly supports, does not support, or is insufficient to judge the proposed relationship, then explain the grounds. A finding that challenges an expectation is still useful if the reasoning is sound. The student's conclusions and evaluation for 9GE0/04 must be their own.
Accuracy, reliability and validity of a claim
Evaluating a conclusion requires identifying how a particular weakness affects the claim, not simply listing things that might have gone wrong.
Accuracy, reliability and validity
Accuracy concerns closeness to the true value; reliability concerns consistency under comparable conditions; validity concerns whether the evidence or method measures what the question requires. Representativeness concerns how well the sample reflects the relevant population, places or times.
These properties are related but distinct. An uncalibrated temperature sensor can give consistently high readings: reliable repetition does not guarantee accuracy. A count of people using a space can be accurately recorded but may not validly measure satisfaction with it. A large sample taken only from daytime visitors can remain unrepresentative of evening users or non-users.
Separate random error, which varies unpredictably and may partly average out across appropriate repeated measurements, from systematic error, which pushes readings in a particular direction. Repeating a measurement with the same biased instrument does not remove systematic error. Sampling bias also survives a larger sample if the same groups remain excluded. For interviews, question wording, social desirability, interviewer influence and selective coding can shape the meaning recorded.
Judge the size and direction of the likely effect where possible. In a practice beach-profile comparison, two endpoint heights are each uncertain by about ±0.1 m. A measured difference of 0.05 m is small relative to that uncertainty, so it gives weak evidence that one profile is higher; a 2 m difference would be less vulnerable to the same uncertainty. Uncertainty does not make every finding worthless, but it changes how firmly it can be stated.
Statistical significance addresses a specified null hypothesis under a test's assumptions. It does not repair biased sampling or invalid measurement, establish a cause, or measure practical importance. A non-significant finding can reflect a small sample or considerable variation. Examine effect size, exceptions and data quality alongside any test outcome. For observational relationships, consider confounding variables that could explain both variables, and the possibility that the supposed effect influences the supposed cause.
Continue this lesson.
You’re previewing NovaLearn without an account. Create a free account to read the rest.
Create free account