Choosing the correct statistical test is a core research-methods skill. In AQA Psychology, students must match an inferential procedure to the form of the data and the way participants were organized.

Why choosing the right test matters

A statistical test is only useful if it fits the data collected in the study. If the wrong test is chosen, the analysis may not suit the structure of the results, and any decision about significance becomes much less meaningful.

In AQA exam questions, students often lose marks by jumping straight to a test name without first identifying the key features of the investigation. Before choosing a test, you should first ask:

• What level of measurement are the data?

• What experimental design produced the data?

These two factors work together. A correct answer usually depends on recognizing both, not just one.

Level of measurement

The level of measurement means the type of data produced by the study. Different statistical tests are designed for different kinds of data, so this is one of the first things to identify.

In AQA Psychology, the main levels of measurement are:

This diagram contrasts nominal and ordinal data by showing that nominal variables are category labels with no ordering, whereas ordinal variables can be placed in a meaningful rank order. It reinforces that ordinal numbers communicate position (higher/lower), not equal numerical distance between ranks. Source

• Nominal data: data sorted into categories or groups

• Ordinal data: data that can be ranked in order

• Interval data: numerical scores where the differences between values are equal and meaningful

This matters because a statistical test must match what the numbers actually represent. Some tests are designed for categories, some for ranks, and some for scores.

Why level of measurement affects test choice

With nominal data, the researcher is dealing with labels, groups, or frequencies. The analysis focuses on how many scores fall into each category.

With ordinal data, the researcher is dealing with order. One score can be higher or lower than another, but the gap between scores is not assumed to be exactly equal. This is common when participants are ranked or when response options only show relative position.

With interval data, the actual distances between scores matter. Because the gaps are equal, the analysis can make fuller use of the numerical information.

A common exam mistake is to assume that any number automatically counts as interval data. That is not always true. If the number only shows order, it should still be treated as ordinal. The key question is not “Are there numbers?” but “What do those numbers mean?”

In question stems, clues about level of measurement often come from the way the results are described:

• if participants are counted in categories, think nominal

• if participants are ranked or given ordered ratings, think ordinal

• if participants produce true scores with equal intervals, think interval

Because statistical tests are built for different levels of data, choosing the wrong level of measurement can lead to the wrong test.

Experimental design

The second major factor is the experimental design.

Here, the important issue is whether the scores are related or unrelated.

For choosing a statistical test, a related design means that the scores in one condition are linked to the scores in another condition. This happens when:

• the same participants take part in both conditions

• participants are matched or paired, so one score is connected to another

An unrelated design means that the scores come from separate participants in each condition. This is typically the case in an independent groups design.

Why experimental design affects test choice

A statistical test must take account of whether the data are paired or independent. Linked scores are treated differently from separate scores because the structure of the comparison is different.

In a related design, the analysis looks at paired results. Each score in one condition has a direct connection to a score in the other condition. In an unrelated design, there is no such link, because the participants in one condition are different from those in the other.

This point is especially important in AQA questions because students sometimes focus on the research-methods label rather than the statistical category:

• Repeated measures and matched pairs are different designs, but for test choice they are both usually treated as related

• Independent groups are treated as unrelated

Two studies may both compare two conditions, but they may still require different tests if one uses related data and the other uses unrelated data.

Using both factors together

The correct choice of statistical test depends on combining both decisions:

• identify the level of measurement

• identify whether the design is related or unrelated

One factor alone is not enough. A test must fit both the form of the data and the structure of the design.

A useful exam routine is:

• read the description of the results carefully

• decide whether the data are nominal, ordinal, or interval

• check whether the same participants, matched participants, or different participants produced the scores

• use both pieces of information to select the test

This method helps prevent rushed answers and makes your reasoning clear.

Common exam mistakes

Several errors appear often in Psychology exams:

• confusing the way data are presented with the level of measurement

• treating matched pairs as unrelated, even though the scores are linked

• assuming that all numerical data must be interval

• choosing a test based only on the topic of the study rather than the data and design

• forgetting that the answer must be justified by the information given in the question

Strong answers are usually brief and precise. Identify the level of measurement, identify whether the design is related or unrelated, and then select the test that matches those features.