Appendix 4: Formulae and statistical tables - Wilcoxon Signed Ranks test process — Edexcel A-Level Psychology

The Wilcoxon Signed Ranks test is a non-parametric statistical test used in Pearson Edexcel A-Level Psychology to analyse differences between two related conditions using ordinal or interval data. For a result to be statistically significant, the calculated value of T must be less than or equal to the critical value found in the statistical tables for the chosen significance level.

The Wilcoxon Signed Ranks test is a core non-parametric statistical tool specified in Appendix 4 of the Pearson Edexcel A-Level Psychology specification. It evaluates whether there is a statistically significant difference between two sets of scores obtained from related data, such as a repeated measures design or matched pairs design, where the data is at least ordinal level.

Mastering this process is essential for Paper 1, Paper 2, and Paper 3 research methods questions. Students must be able to justify the choice of test, calculate the test statistic T by ranking signed differences, handle tied ranks and zero differences, and accurately interpret critical value tables to accept or reject hypotheses.

• →Test criteria: Used for a test of difference, a repeated measures or matched pairs design (related data), and ordinal or interval/ratio data.
• →Zero difference exclusion: Any participant scoring identically across both conditions (difference = 0) is excluded, which directly reduces the value of N used for critical tables.
• →Ranking absolute values: Differences are ordered from smallest to largest irrespective of mathematical sign (+ or -), with tied scores assigned the mean of the ranks they span.
• →Test statistic T: Calculated T is defined as the smaller sum of the assigned signed ranks (T = min(sum R+, sum R-)).
• →Significance threshold rule: For the Wilcoxon test, the observed value of T must be less than or equal to the critical value to achieve significance at the designated alpha level.

• State the design first: paired values only exist because the same participants, or matched pairs, were tested under both conditions.
• Work in a consistent direction, for example condition A minus condition B on every row, recording a minus sign wherever the second value is the larger.
• Keep the sign in the difference column, since it is what separates the two rank totals later; only the ranking stage treats a difference purely as a size.
• Justify the test as well as performing it: ordinal or skewed interval data, a repeated measures design, and a hypothesis predicting a difference rather than an association.
• Rank the size of each difference only; the sign is set aside here and recovered when the ranks are sorted into totals.
• The smallest difference takes the first position and ranking runs upwards, so the largest difference carries the highest position.
• Share tied positions: give each tied difference the mean of the positions they would have occupied, then resume at the next unused position.
• Explain why ranking is used at all: positions can be assigned to scores that are ordered but not equally spaced, which is what lets ordinal data into the test.
• A row showing no change is discarded before ranking, because a participant who scored the same twice offers no evidence for either direction of change.
• The sample size used at the end counts only participants who produced a non zero difference, so recalculate it after the discards instead of using the number of people tested.
• Positions are awarded to the size of a difference, so minus four and plus four compete for the same position; the signs return when the positions are sorted into two totals.
• Explain the consequence for the decision: discarding ties lowers the sample size, the critical value drops with it, and the test becomes more conservative.
• Group the positions by the sign of the difference each came from, then total those belonging to the rows above zero.
• Say what a positive difference means in this particular table, since that depends entirely on which condition was subtracted from which.
• Check the arithmetic: the two totals together must equal the sum of all positions awarded, which is the number of positions multiplied by one more than that number, then halved.
• Interpret the pair of totals: a large imbalance between them is what a significant result reflects, whereas nearly equal totals mean the conditions barely differed.
• Total the positions attached to the rows whose difference carried a minus sign, keeping the discarded unchanged rows out of both totals.
• Relate the pair of totals to the hypothesis: a small total on one side against a large one on the other means most participants changed in the same direction.
• Report both totals, because the statistic is defined by comparing them and a single total on its own cannot be checked by anybody reading the report.
• Recognise that the data being added here are ordinal, since it is positions rather than raw magnitudes that are summed.
• Identify the statistic as the smaller of the two rank totals and state its value plainly before comparing it with anything.
• Compare it with the critical value taken from the table at the adjusted sample size, for the correct number of tails and at the five per cent significance level.
• Apply the comparison in the right direction: the result counts as significant only where the calculated value is equal to or smaller than the critical value.
• Write the conclusion in the terms of the study, rejecting the null hypothesis and stating the direction of the difference, or retaining it, and note that the probability of a type one error was held at five per cent.
• Define the figure as the count of non zero differences rather than the number of people who took part in the study.
• Use it to enter the table of critical values: this figure picks the row, while the number of tails and the significance level pick the column.
• Report it with the statistic so the comparison can be checked, for example a calculated value of six against a critical value of eight for ten usable differences, two tailed.
• Comment on power where the question invites evaluation, since a small number of usable differences makes a significant result unlikely even when a real effect exists.
• Credit comes from naming the test and justifying it in the same breath: a test of difference, a related design of repeated measures or matched pairs, and data that are ordinal or can be ranked.
• Say that the figure used to enter the table is the number of pairs that produced a non-zero difference rather than the number of participants tested, and state the significance level and whether a one tailed or two tailed value is being read.
• State the decision in the examiner's language: because the observed statistic is smaller than or the same as the tabled figure at five per cent, the null hypothesis is rejected and the difference is described as significant.
• A mark is available for the direction rule itself, that Wilcoxon and Mann-Whitney are the tests where a small observed figure signals significance while chi-squared, Spearman and the t tests need a large one.

• 💡The skills paper sets this as a partly completed table and asks for the missing values, so show the subtraction rather than only the answer.
• 💡Calculation items are often paired with a short justification of the test, so learn the design, the level of measurement and the hypothesis type as one package.
• 💡Read the stem for one tailed or two tailed wording, because that choice decides which column of the table of critical values you will need at the end.
• 💡The paper usually supplies a table with one or two positions missing, so practise finishing a ranking under time rather than starting one from scratch.
• 💡Show the ranks in a column beside the differences, because a marker awarding method marks has to see where each number came from.
• 💡If the data contain ties, mention them in an explain answer, since tied ranks are a recognised limitation of rank based tests.
• 💡Stems are written with one or two unchanged pairs in the table on purpose, so scan that column before starting to rank.
• 💡A short explain item often asks why a participant was excluded; answer in terms of the absence of a direction of change, not by reciting the rule.
• 💡Quote the sample size you finally used beside the statistic, because the comparison with the critical value cannot be credited without it.
• 💡Show both totals even where the question asks only for the final statistic, because method marks are available for the two sums.
• 💡Use the arithmetic check before moving on, since a total that fails it points to a ranking error that would otherwise carry through to the conclusion.
• 💡In an explain answer, link the larger total to the direction stated in the hypothesis rather than simply quoting the figure.
• 💡Lay the working out as a labelled column, because a marker credits a correctly totalled column even where the final comparison goes astray.
• 💡Where the table is already part filled, check the given positions against the differences before totalling, since questions sometimes plant an error for you to spot.
• 💡An explain item may ask what a very small total indicates, so be ready to write about consistency of change across participants.
• 💡The tables are printed in the paper, so practise reading the correct row and column at speed rather than trying to memorise values.
• 💡A frequent three or four mark item supplies the calculated value and the table and asks you to state and justify the conclusion, so always finish with a sentence about the hypothesis.
• 💡Where the stem gives a directional hypothesis, use the one tailed column and say in your answer why you did.
• 💡Count the usable differences as you cross off the unchanged pairs; the arithmetic in these items is easy and it is the bookkeeping that loses marks.
• 💡Questions asking you to evaluate a study's statistics reward a comment on sample size and power, so keep one sentence prepared.
• 💡State the figure in your concluding sentence together with the number of tails and the level, in the format the mark scheme models.
• 💡Questions usually supply a small table of before and after scores and ask you to calculate the statistic, so practise laying out columns for difference, sign and rank and show the working, because method marks survive an arithmetic slip.
• 💡Another frequent version gives the observed figure and the critical value and asks what the comparison shows; answer with the comparison, then the decision about the null hypothesis, then the level of significance, in that order.
• 💡Be ready to justify the choice of test against Mann-Whitney U, since the only thing separating them in most stems is whether the design is related or independent groups.
• 💡Use of inferential statistics is nomothetic and reductionist, so an evaluation question can be turned towards what a group level probability says about any individual participant.
• 💡Always state the three justification criteria explicitly when asked why Wilcoxon was chosen: 'difference', 'repeated measures (or matched pairs)', and 'at least ordinal data'.
• 💡Show full workings when summing positive and negative ranks; writing out 'Sum of + ranks = ...' and 'Sum of - ranks = ...' secures working marks even if a minor arithmetic slip occurs.
• 💡When concluding significance, quote all four elements: the calculated T value, the critical value, the N value, and the p level (including one-tailed vs two-tailed).

• Always taking the smaller value from the larger, so every difference comes out positive and one of the two rank totals ends up empty.
• Dropping the minus signs at this stage because a later instruction mentions ignoring them, when they are ignored only while ranking.
• Choosing this test for an independent groups design, where Mann-Whitney is the test that actually fits.
• Ranking from largest downwards, which reverses every position and produces the wrong statistic while the arithmetic still looks tidy.
• Ranking signed values, so every negative difference sits below every positive one and the two totals become meaningless.
• Giving two tied differences consecutive positions instead of the shared mean, which shifts every position that follows and corrupts both totals.
• Ranking the unchanged rows as though they were the smallest differences, which lifts every other position by one and alters both totals.
• Dropping the unchanged rows from the ranking but still counting them in the sample size, so the wrong row of the table of critical values is read.
• Reading the instruction about signs as permission to delete them, and then being unable to separate the positive from the negative total.
• Adding the differences rather than their positions, which yields a number that looks plausible but cannot be compared with any critical value.
• Including the discarded unchanged rows in the total, or counting a shared position twice after a tie.
• Reporting the total without saying which condition scored higher, so the result cannot be tied back to the hypothesis.
• Carrying the minus signs into the total and producing a negative number, when it is the positions and not the differences that are being added.
• Assuming this total is always the one used, rather than whichever of the two totals turns out to be smaller.
• Counting a shared position in both totals after a tie, so the two sums exceed the total that the arithmetic check predicts.
• Taking the larger total as the statistic, which nearly always exceeds the critical value and produces a confident but wrong conclusion of no difference.
• Importing the greater than rule from chi-squared, so a genuinely significant result is reported as non significant.
• Reporting significance without naming the level, the number of tails or the sample size, all of which the mark scheme expects alongside the decision itself.
• Using the original number of participants, which sends you to the wrong row and usually to a critical value that is far too generous.
• Confusing this count with the number of conditions, or with degrees of freedom, which belongs to other tests entirely.
• Treating a non significant result from a handful of usable differences as proof that the conditions do not differ, rather than as a study with too little power to tell.
• Keeping pairs whose difference is zero in the count, which inflates the sample size, sends the student to the wrong row of the table and often flips a significant result into a non-significant one.
• Ranking the differences with their plus and minus signs attached, or ranking the raw scores instead of the differences, so the rank totals are already meaningless before the table is opened.
• Taking the larger of the two rank totals, or importing the greater than rule from chi-squared, so a genuine effect is written up as no effect.
• Writing that the result proves the hypothesis. Examiners want the null hypothesis rejected, with a five per cent probability that the result is due to chance, not proof.
• Retaining zero differences in the sample size: Students frequently leave difference scores of zero in the ranking process or keep the original N when consulting the statistical table, resulting in the selection of the wrong critical value.
• Ranking including the sign: Students occasionally rank negative differences as smaller than positive differences (e.g. treating -5 as smaller than +1) rather than ranking the absolute magnitude of the differences.
• Confusing the decision rule: Many candidates default to the assumption that calculated values must always exceed critical values, incorrectly concluding that higher T values indicate statistical significance.

1. 1Week 1 (Day 1-2): Master the three justification conditions for Wilcoxon versus Mann-Whitney U, Sign test, and related t-test.
2. 2Week 1 (Day 3-5): Practice calculating difference scores, omitting zero scores, and resolving tied ranks with fractional values across small mock datasets.
3. 3Week 2 (Day 1-3): Practice summing positive and negative ranks to derive calculated T, then using Pearson Edexcel statistical tables to identify correct critical values.
4. 4Week 2 (Day 4-5): Write full analytical conclusions linking calculated T, critical values, alpha levels, and contextual hypothesis acceptance or rejection.

• 📋Short-answer calculation (2-4 marks): Calculating difference scores, ranking, and identifying T from a provided data table.
• 📋Test justification (2-3 marks): Explaining why the Wilcoxon Signed Ranks test is appropriate for a described empirical scenario.
• 📋Critical table interpretation (3-4 marks): Identifying the critical value given specific parameters (N, tails, p-value) and writing a formal statistical conclusion.