D.1 – Handling data — Edexcel A-Level Psychology

Handling data in Pearson Edexcel A-Level Psychology involves calculating and interpreting descriptive statistics, levels of measurement, measures of central tendency, dispersion, and graphical representations. Mastering these quantitative skills ensures students can rigorously analyse raw psychological research data and draw valid scientific conclusions.

Handling data covers the fundamental quantitative techniques required to collect, summarize, present, and evaluate empirical evidence in psychological investigations. Students learn to classify raw data into levels of measurement, calculate appropriate measures of central tendency and dispersion, and construct accurate graphical displays.

This mathematical toolkit underpins research methods across all topic areas in Pearson Edexcel Psychology, including Social, Cognitive, Biological, and Learning theories. Developing data literacy allows students to assess the validity of empirical findings, spot researcher bias or data distortion, and earn crucial quantitative marks across Papers 1, 2, and 3.

• →Levels of Measurement: Nominal (categorical), Ordinal (ordered ranks without equal intervals), and Interval/Ratio (standardised units of equal scale).
• →Measures of Central Tendency: Mean (arithmetic average sensitive to all scores), Median (middle score resistant to skew), and Mode (most frequent value for nominal sets).
• →Measures of Dispersion: Range (spread between top and bottom values) and Standard Deviation (average dispersion of scores around the mean).
• →Data Visualisation: Accurately constructing bar charts, histograms, scattergrams, and frequency polygons with labelled axes, units, and precise scales.

• The rounded value needs the correct number of digits written out, so a coefficient of 0.80 keeps its trailing zero rather than appearing as 0.8.
• Interpretation is credited alongside the arithmetic, so state strength and direction, such as a strong positive relationship, and name both variables.
• Where a decision is asked for, compare the calculated coefficient with the critical value for the number of pairs and state whether the result is significant.
• Name the test that produced the coefficient, Spearman's rho for ordinal data or Pearson's product moment for interval data, because the justification often carries the next mark.
• Write the total and the division before the answer, because the method mark is for sum of scores over number of scores.
• Give the answer in the units of the original data and to sensible accuracy, usually one decimal place beyond the raw scores.
• Comparison earns credit when the direction is stated in the study's own terms, for example that participants recalled more words in the imagery condition than in the rehearsal condition.
• Justifying the choice, that the data are interval and free of extreme scores, is a standard two mark answer and is asked as often as the calculation.
• Credit is given for the correct display plus the reason for it: a bar chart because the data are in discrete named categories, a histogram because the scores are continuous and grouped into equal class intervals.
• Markers look for the conventions on the sketch itself, namely gaps between bars on a bar chart, touching bars on a histogram, the frequency on the vertical axis and a suitable scale that begins at zero.
• Both axes must be labelled with the variable and, where relevant, the units; a title naming the two conditions compared, such as the mean speed estimate for each verb in Loftus and Palmer's leading question study, secures the presentation mark.
• Where the question asks for interpretation, say what the shape shows, for example that the tallest bar identifies the modal category or that the histogram is skewed towards the lower class intervals.
• Define the level as the probability that the result occurred by chance, so at the five per cent level there is a one in twenty chance of wrongly rejecting a true null hypothesis.
• Name the error each level controls: the more stringent level reduces the risk of a type one error, a false positive, but increases the risk of a type two error, a false negative in which a genuine effect is missed.
• Give a reason for choosing the stricter level, such as research with clinical or medical consequences, a replication testing a contested claim, or a finding that would overturn accepted theory.
• Credit follows an applied statement, for example that a researcher testing a new therapy uses the stricter level because recommending an ineffective treatment would harm patients.
• Marks come from a procedure described in order: define the target population, obtain the sampling frame, then state the selection mechanism, and for a stratified sample the calculation of proportions before random selection within each stratum.
• Say explicitly that in a random sample every member of the target population has an equal chance of being selected, and that in a stratified sample selection within each stratum is itself random.
• Use the numbers in the stem, so a population of two hundred containing one hundred and twenty women sampled at ten per cent yields twelve women and eight men.
• Evaluation credit follows a named limitation tied to generalisability, such as Milgram's newspaper advertisement producing a volunteer sample of forty American men, which limits how far obedience rates transfer to women or to other cultures.
• A calculation mark needs the method shown, so write the total and the divisor for a mean, and state that the scores were placed in order before identifying the middle value for a median.
• Justify the choice by the data, for example that a mean is appropriate for the speed estimates in Loftus and Palmer's 1974 study, where the smashed condition averaged 40.8 miles per hour against 31.8 for contacted, because those are interval measurements with no extreme outliers.
• Say what each measure ignores: the median ignores the size of the scores either side of it, the mode ignores everything except frequency, and the mean is the only one that uses all the data.
• Pair the measure of central tendency with a measure of dispersion, since examiners reward the point that a mean alone hides how consistent the participants were.
• Working is credited step by step, so show the mean, the squared deviations, the sum of those squares, the division and the final root rather than only the answer.
• Interpretation earns as much as arithmetic: state that the condition with the smaller value produced more consistent performance and the larger value shows greater individual differences.
• Say why it is preferred to the range, namely that the range uses only the highest and lowest scores while this uses every score in the set, making it the more sensitive measure.
• Note the requirement for interval data and the effect of an outlier, since squaring the deviations means one extreme score inflates the figure disproportionately.
• The sketch is credited for labelling each axis with one co-variable and for a plotted pattern that clearly matches the type named, together with a title stating the two measures.
• Name the pattern precisely, so a downward trend is a negative correlation rather than a weak or poor one, and an absence of any trend is stated as no correlation rather than as a failed study.
• Distinguish strength from direction, since a strongly negative relationship is as strong as a strongly positive one and only the sign differs.
• Evaluation credit comes from the third variable problem and the absence of manipulation, for example that a relationship between hours of revision and grades could be driven by motivation affecting both.
• The choice of a statistical test depends on three factors: the type of hypothesis (difference or correlation), the experimental design (independent or repeated measures), and the level of measurement (nominal, ordinal, or interval).
• In the Wilcoxon signed ranks test, the calculated value (T) is found by calculating the difference for each pair, discarding zero differences, ranking the absolute differences, and taking the smaller of the two rank sums (positive or negative).
• For the Wilcoxon, Mann-Whitney, and sign tests, the calculated value must be equal to or less than the critical value to be significant, whereas for chi-squared and Spearman's rho, the calculated value (using the absolute value for negative Spearman correlations) must be equal to or greater.
• Interpreting a statistical test requires comparing the calculated value to a critical value, which is determined by the sample size (N) or degrees of freedom (df), the significance level (e.g., p = 0.05), and whether the hypothesis is one- or two-tailed.
• Credit follows the rounded figures being written down, so show the rounded total, the rounded number of participants and the division rather than presenting a bare estimate.
• State that the average is the overall total divided by the number of participants, and name what the divisor is in the scenario rather than assuming it.
• An answer given to a sensible degree of precision is rewarded, so an estimate is expressed to one significant figure or to the nearest ten, not to several decimal places.
• Where the question asks whether a reported figure is plausible, say which way the estimate falls and by roughly what factor the reported figure differs.
• Identify nominal data when a study categorises participants and records the frequency of occurrences within each distinct, non-overlapping category.
• Identify ordinal data when a study uses ranked positions or subjective rating scales where the intervals between points cannot be proven to be mathematically equal.
• Identify interval data when a study measures variables using fixed, standardised units with genuinely equal intervals, such as time in seconds or temperature.
• Distinguish between the levels of measurement to determine the appropriate descriptive statistics, such as using the mode for nominal data, median for ordinal, and mean for interval data.
• Credit for saying that in a normal distribution the mean, the median and the mode fall at the same point under the peak of a symmetrical curve, with most scores clustered near that centre.
• Credit for placing the mean nearest the tail: reading left to right, a positive skew gives mode, median, mean, and a negative skew gives mean, median, mode.
• Credit for naming the direction of the tail and its cause, such as a small number of unusually slow responses producing a long right hand tail in timed data.
• Credit for linking shape to the choice of average, saying that the median better represents a skewed set because the mean is pulled towards the extreme scores.
• Credit for linking normality to test choice, since a parametric test such as the related t test assumes interval data drawn from a roughly normal distribution.
• Credit for naming the test and then justifying it against all three criteria: the kind of hypothesis, the design, and the level of measurement.
• Credit for identifying the design from the procedure, so that two scores from every participant is repeated measures and needs a related test, while one score from each of two separate groups is independent groups and needs an unrelated test.
• Credit for reading the level of measurement off the way the data were gathered, such as a tally of how many people chose each answer being nominal and time in seconds being interval.
• Credit for saying that a t test is parametric and requires interval data, an approximately normal distribution and similar variances, so an ordinal equivalent is used when those conditions fail.
• Credit for treating matched pairs as a related design, because matching pairs of participants on relevant variables performs the same job as using the same person twice.
• Credit for stating the comparison in the correct direction for the named test, and for knowing that Mann Whitney U and Wilcoxon are the two where the calculated value has to be equal to or less than the critical value.
• Credit for using the one tailed column because the hypothesis is directional, and for naming the level as the five percent level of probability.
• Credit for a conclusion written in terms of the hypotheses, rejecting the null hypothesis and accepting the alternative hypothesis because the probability of the result arising by chance is below five percent.
• Credit for checking the stricter one percent column once a result has passed at five percent, if a stronger claim about the finding is being made.
• Credit for reading the correct row, using degrees of freedom for chi square and the t tests and the sample or pair count for the rank based tests.
• Credit for defining the range as the difference between the highest and the lowest score, and the standard deviation as the average amount by which scores differ from their own mean.
• Credit for the comparison the question wants, namely that the range is fixed by only two scores so one outlier distorts it completely, while the standard deviation is calculated from every score and therefore represents the spread of the whole set.
• Credit for a point made on the data supplied, such as noting that deleting the outlying score would shrink the range dramatically but alter the standard deviation only a little.
• Credit for interpreting a large standard deviation as showing that the mean is less representative of the group and that individual differences within that condition are greater.
• Credit for linking dispersion to comparison between conditions, since two conditions can share a mean and still differ in how consistent the participants were.
• Credit for a contrast that names the method as well as the data, such as an open ended interview question producing words and a rating scale producing scores.
• Credit for setting out the conversion in steps: decide the coding categories, work through the transcript, tally each occurrence, and report the totals as frequencies.
• Credit for naming inter rater reliability as the check on that coding, with a second researcher coding the same transcript and the two sets of tallies compared or correlated.
• Credit for an example from the specification's own studies, such as the observers in Bandura, Ross and Ross recording aggressive acts in short timed intervals so that behaviour became a score, or Rosenhan turning pseudopatient ward notes into counts of days admitted and tablets dispensed.
• Credit for evaluation on both sides, since counting allows statistical testing and improves objectivity but discards the context that made the words worth gathering.
• Credit for applying the rule to the scenario, so that anything gathered by a researcher dealing directly with participants is primary data whatever form it takes.
• Credit for naming a genuine secondary source, such as official crime or health statistics, medical records, school records, or the data sets pooled in a meta analysis.
• Credit for one strength and one weakness on each side, for instance that primary data are tailored to the aim but expensive, while secondary data are quick and plentiful but of unknown quality.
• Credit for the meta analysis link, such as van IJzendoorn and Kroonenberg pooling Strange Situation findings from many countries, which was only possible because the original data already existed.
• Credit for noticing that one study can use both, as when brain scans are collected first hand while the participants' offence histories are taken from court records.

• 💡The rounding instruction usually sits at the end of the calculate stem, so read the whole sentence before writing the answer.
• 💡A coefficient close to zero still needs interpreting; say there is little or no relationship rather than saying the study failed.
• 💡Remember the critical value rule changes direction by test, since Spearman and chi squared need the calculated value to be equal to or larger, while Wilcoxon and Mann Whitney need it equal to or smaller.
• 💡Class experiment stems give raw data in a table, so work down each column and write the total before dividing, since that total is often worth a mark on its own.
• 💡The follow up nearly always asks for an interpretation or a justification of the measure, so be ready to explain the choice against the median and the mode.
• 💡Check whether the question wants the average for each condition or the difference between them, because the mark scheme will not accept one for the other.
• 💡This appears as a short sketching task worth two to four marks with a small table of data in the stem, so use the stem's own category names and frequencies rather than inventing numbers.
• 💡A frequent pairing asks you to select the display and then justify it, so write one sentence naming the level of measurement of the data before describing the graph.
• 💡Draw the axes and label them before plotting anything, because the labelling marks are the ones candidates most often forfeit by running out of time.
• 💡This is usually two or three marks asking why a researcher chose one level rather than the other, so answer with the risk being controlled and the consequence in the scenario given.
• 💡Learn the standard conclusion wording, that the probability the result is due to chance is equal to or less than the stated level, because examiners reward the phrasing as well as the idea.
• 💡The same material is examined inside a test question, where you state the level and the number of tails before reading the critical value from the table.
• 💡The stem gives a population and a required sample size, so show the arithmetic for each stratum rather than simply asserting the sample is proportional.
• 💡Questions pair the method with one strength and one weakness, so prepare a practicality point for random sampling and a time cost point for stratified sampling.
• 💡In extended answers, link the sampling technique to population validity and to cultural bias, since studies using only one nation or one sex invite the ethnocentrism and androcentrism criticism.
• 💡Calculations are set from a table in the stem, so copy the figures out once and check the count before dividing; answers are usually required to two decimal places unless told otherwise.
• 💡The follow up almost always asks why that measure is appropriate, so have one sentence ready linking the level of measurement and the presence or absence of outliers.
• 💡Where a distribution is described as skewed, the expected answer is the median, and saying so explicitly earns the justification mark.
• 💡Calculators are allowed, so most marks go to interpreting two figures in a results table, for example a value of 2.4 in one condition against 7.9 in the other.
• 💡Where a question gives the mean and this measure together, comment on both, because a conclusion that ignores the spread cannot be a top band answer.
• 💡Link it to the choice of inferential test by stating that it requires interval level data, which is the same condition a parametric test needs.
• 💡Sketching questions carry a mark for each labelled axis and a mark for the pattern, so label first and plot roughly eight to ten points rather than three.
• 💡The standard follow up asks why the researcher cannot conclude that one variable caused the other, so keep a named third variable ready for the scenario in the stem.
• 💡If asked which test analyses the data, say Spearman's rho and justify it by the relationship being tested and the ordinal level of the scores.
• 💡Learn the statistical test selection criteria as a grid and reproduce it in rough before answering questions that require you to justify a test choice.
• 💡When provided with a calculated value and a table of critical values, quote both figures in your answer and state the comparison explicitly rather than just giving the final verdict.
• 💡Where tied scores appear in a ranking calculation, ensure you calculate the shared rank as the mean of the positions they occupy, as this is a key step in the method.
• 💡Questions are worded as an instruction to estimate, so an exact calculation is not required and rounding shown in the working is safer than an unexplained exact figure.
• 💡Use the technique as a free check on every other calculation in the paper, since a misplaced decimal point is the commonest way to lose marks that were otherwise earned.
• 💡If the stem gives a total and a sample size in unfriendly numbers, round both in the same direction consciously and say which way the estimate is likely to err.
• 💡When identifying the level of measurement, always provide a justification based on the specific data characteristics described in the study.
• 💡Remember that the level of measurement must be correctly identified before selecting the appropriate inferential statistical test.
• 💡Where a study reports both a count of participants and a measured score, specify which level of measurement applies to which piece of data rather than giving one label for the whole study.
• 💡These items often print a skewed curve with three unlabelled arrows and ask which is the mean, so find the tail first and then read mode, median and mean in the direction of that tail.
• 💡If given a set of scores, comment on the shape as well as the average and say what a single outlying score does to the mean.
• 💡The command word here is usually describe or explain for two or three marks, so justify the position you have marked in one clause rather than redrawing the graph.
• 💡The question gives a short scenario and asks you to explain why a named test is appropriate, so write one sentence per criterion and tie each sentence to a detail printed in the scenario.
• 💡When the stem says ranks, positions or a rating out of ten, expect Mann Whitney U, Wilcoxon or Spearman's rho rather than a t test.
• 💡Extended research methods questions that ask you to design a study almost always want the test and its justification, so leave a line for it at the end of your plan.
• 💡The paper prints an extract from the tables rather than the whole set, so the marks are for choosing the right cell and stating the comparison, not for recalling numbers.
• 💡Write the calculated value, the critical value, the participant count or degrees of freedom, the tail and the level in one sentence, then give the decision about the null hypothesis in the next.
• 💡If asked to explain why a result is significant, name the direction of the comparison, because saying only that the value is bigger earns nothing on a Wilcoxon or Mann Whitney item.
• 💡The item usually prints a small table containing an obvious outlier and awards marks for saying what that score does to each measure, so quote the actual figures rather than talking about dispersion in general.
• 💡Pair every mean you quote in an evaluation with its standard deviation and say whether the conditions overlap.
• 💡This is often a two or three mark explain question, so give the definition in a clause and spend the remaining lines on the effect of the outlying score.
• 💡The question normally supplies a short transcript and asks how it could be made numerical, so name a category you would code and say how you would count it in that extract.
• 💡When evaluating, link the data type directly to validity and to reliability rather than listing untethered advantages.
• 💡Expect a follow up asking which test could then be used, and be ready to say chi square because coded tallies are frequencies in categories.
• 💡A single mark item may simply ask which type a described data set is, so give the word and a short reason naming who collected it.
• 💡In longer methods answers use this distinction to evaluate meta analyses and official statistics, noting that official figures under record events that go unreported.
• 💡If the stem mentions records or archives, expect a follow up asking for one limitation of that source, so have the unknown procedure point ready.
• 💡When asked to calculate percentages or ratios, always show complete written working even for straightforward steps to secure method marks.
• 💡Structure data interpretation questions using a comparative formula: compare the mean/median first, compare dispersion second, and link both explicitly back to the operationalised psychological variables.
• 💡Always state whether high or low values denote better or worse performance in the context of the study (e.g. higher reaction time scores mean slower responses).

• Counting decimal places instead of figures, which turns a coefficient of 0.0432 into 0.04 and loses a whole order of precision.
• Dropping the minus sign and then describing a negative correlation as evidence that both variables rise together.
• Rounding at every stage of a longer calculation so the final value drifts, instead of rounding once at the end.
• Quoting a coefficient to five or six digits when the raw scores were whole numbers, which claims precision the data cannot support.
• Dividing by the number of conditions, or by the size of one condition after pooling both, instead of by the number of scores actually added.
• Including a participant who withdrew, or treating a blank cell as a score of zero, which drags the figure down.
• Reporting the difference between two conditions as a result in itself and forgetting that it needs a test of significance.
• Calculating an average from nominal data such as favourite category, where only the mode carries meaning.
• Drawing a histogram with gaps between the bars, or a bar chart with the bars touching; the gap is the visual signal that the horizontal axis is not a continuous scale, and losing it loses the mark.
• Plotting each participant's raw score along the horizontal axis instead of the categories or class intervals, which produces a graph of individuals rather than a frequency distribution.
• Leaving the axes unlabelled or unscaled, which examiners treat as an incomplete sketch however neat the bars are.
• Using a histogram for data in named categories such as preferred hand or gender, where there is no underlying continuum to divide into classes.
• Writing that the results are ninety five per cent certain to be true, or that the hypothesis has a ninety five per cent chance of being correct; the probability attaches to the data under the null hypothesis, not to the truth of the theory.
• Describing the stricter level as more accurate or more scientific, when it simply shifts the risk from one type of error to the other.
• Swapping type one and type two, so that a claimed effect which is not real is labelled a type two error.
• Saying a result was not significant so the null hypothesis is proved true, rather than saying it is retained.
• Describing an opportunity sample, people who happened to be available, and labelling it random; randomness is a property of the selection mechanism, not of who turns up.
• Stopping after identifying the strata and their proportions, without stating that participants are then chosen randomly within each stratum, which is the step that carries the final mark.
• Claiming a random sample guarantees representativeness, when a small random sample can easily be skewed by chance alone.
• Giving strata figures that do not add up to the sample size required by the stem, which costs the calculation mark even when the method is right.
• Finding the median without first putting the scores in order, which produces whichever value happens to sit in the middle of the list as written.
• Reporting a mean for data in named categories, such as favourite colour or yes and no answers, where only the mode is meaningful.
• Confusing the mode with the middle value or with the highest frequency count itself, rather than the score that occurs most often.
• Dividing by the number of conditions or trials instead of the number of scores, which gives an answer that is out by a whole factor.
• Forgetting the square root at the end, which leaves the variance and an answer roughly the square of the one wanted.
• Squaring the sum of the deviations instead of squaring each deviation first, which gives zero or near zero because the deviations cancel.
• Reading a large value as a bad or invalid result, when it is simply a statement about spread and may itself be the interesting finding.
• Reporting a negative value, which is impossible once deviations have been squared, and usually signals a sign error carried through the working.
• Putting an experimental condition on the horizontal axis and a score on the vertical axis, which is a bar chart situation and not a correlation at all.
• Joining the plotted points together with a line, as in a time series graph, instead of leaving them as separate points with at most a line of best fit.
• Concluding that one co-variable caused the other, which no correlational design can support however strong the pattern looks.
• Reading a near zero coefficient as proof of no relationship, when a curvilinear pattern such as the inverted shape reported by Yerkes and Dodson for arousal and performance gives a low linear value despite a real association.
• Ranking each group separately in the Mann-Whitney U test, when all scores from both groups must be ranked together as a single set before the ranks are totalled by group.
• Counting zero differences in the sample size (N) for the Wilcoxon or sign test; participants with no difference must be dropped, reducing the sample size accordingly.
• Concluding significance without checking the direction of the test or the significance level, such as incorrectly using a critical value for a one-tailed test when the hypothesis is two-tailed.
• Dividing by the number of trials, items or conditions instead of the number of participants, which produces an estimate that is out by a whole order of magnitude.
• Misplacing the decimal point and then accepting the answer because the calculator produced it, which is precisely the error this technique exists to catch.
• Quoting an estimate to several decimal places, which wastes time and misrepresents a rough calculation as a precise one.
• Treating the estimate as the actual result in the conclusion, rather than as a check on the figure that was properly calculated.
• Calling a rating scale interval simply because it uses numbers, when the distance between adjacent points is not a fixed unit and the data are ordinal.
• Calling reaction time ordinal on the grounds that the times can be put in order, when seconds are equal standardised units and the data are interval.
• Confusing the level of measurement with the qualitative and quantitative distinction, or with the independent and repeated measures design, which are separate decisions.
• Assuming nominal means no numbers are involved, when nominal data are counts of how many participants fall into each category.
• Naming the skew from where the bulk of the scores sits rather than from the direction of the tail, so a set of mostly low scores with a few very high ones gets called negatively skewed when it is positively skewed.
• Marking the mean at the highest point of a skewed curve. Only the mode sits under the peak once the curve is asymmetrical.
• Assuming a large sample must give a normal distribution, when sample size does nothing to remove a floor or ceiling effect.
• Describing the curve without ever mentioning the raw scores the question printed, which loses the application marks.
• Choosing chi square because the data contain numbers. Chi square works only on frequency counts of people falling into categories, not on each participant's individual score.
• Calling a study a correlation because it has two sets of numbers, when a correlation needs two measured variables from the same people and no manipulated independent variable.
• Analysing matched pairs with an unrelated test, which throws away the control the matching was designed to give.
• Naming a test without mentioning the data at all, which leaves the explain marks unearned even when the test is right.
• Using the two tailed column for a directional hypothesis, which sets a harder critical value and can turn a significant finding into a non significant one.
• Entering the chi square table with the number of participants instead of the degrees of freedom, so the wrong row is read entirely.
• Writing that the result proves the hypothesis, when an inferential test only supports or fails to support it at a stated level of probability.
• Treating a non significant result as proof that no effect exists, rather than as a failure to reject the null hypothesis, which is where a type two error hides.
• Describing the standard deviation as the difference between the means of two conditions, when it describes the spread inside a single set of scores.
• Claiming the standard deviation is unaffected by outliers. It is less affected than the range, and that relative advantage is the answer, not an absolute immunity.
• Quoting means alone when asked what the data show, so the answer never says whether the two sets of scores overlap.
• Calling the range more useful because it is easier to calculate, which misses that ease of calculation is not what the question rewards here.
• Calling qualitative data unscientific, when they are analysed systematically through content or thematic analysis and can be checked for reliability.
• Describing content analysis only as turning words into numbers, without naming the coding categories or the tallying, which is exactly where the marks are.
• Confusing the data with the method, since a single questionnaire yields words from its open items and scores from its closed ones.
• Writing a generic list of strengths that is never tied to the interview extract the question printed.
• Assuming secondary data must be words or come from textbooks, when a government statistics table is quantitative secondary data.
• Calling the results of a previously published study primary data because they are real research. They are primary to the team who collected them and secondary to anyone reusing them.
• Giving only definitions when the question asks for an evaluation of using existing records in a named investigation.
• Saying secondary data are unreliable with no reason, instead of pointing to unchecked procedures, changed definitions or under reporting.
• Believing the mean is always the best measure of central tendency; it is heavily distorted by extreme values or skewed data, making the median more valid in those scenarios.
• Treating rating scales (e.g. Likert scales from 1 to 5) as interval data; psychological scales typically provide ordinal data because intervals between subjective ratings cannot be proven equal.
• Forgetting to label graph axes fully with both the variable name and units of measurement, resulting in an automatic loss of presentation marks.

1. 1Day 1-3: Master definitions and classification of Nominal, Ordinal, and Interval/Ratio data using examples from Cognitive and Social core studies.
2. 2Day 4-6: Practice manual calculations of mean, median, mode, and range, followed by drilling the advantages and limitations of each measure.
3. 3Day 7-9: Learn standard deviation interpretation and practice constructing and labelling bar charts, histograms, and scattergrams from raw data sets.
4. 4Day 10-14: Complete Pearson Edexcel past paper research methods sections, focusing on 3-mark and 4-mark data analysis and data justification stems.

• 📋Mathematical calculation questions (1-3 marks): Requiring direct calculation of percentages, fractions, mean, median, or range from a table of participant scores.
• 📋Justification questions (2-4 marks): Asking why a researcher chose a specific graph, average, or measure of dispersion based on the data characteristics provided.
• 📋Data analysis and conclusion questions (4-6 marks): Interpreting tables of descriptive statistics to state findings and evaluate the consistency of participant behaviour.