Psychology Dissertation Descriptive Statistics

Psychology dissertation descriptive statistics summarise who participated, what was measured, and how observations were distributed before any inferential claim is made. Good description is not a decorative prelude to hypothesis testing. It is the foundation for detecting errors, choosing suitable analyses, interpreting effects, and allowing readers to understand the sample.

This guide explains how to select means, medians, standard deviations, interquartile ranges, counts, proportions, ranges, and distribution plots for psychology research. It also covers repeated measures, scale scores, missing data, demographic tables, subgroup summaries, APA-style presentation, and the mistakes that make descriptive tables misleading.

What are Psychology Dissertation Descriptive Statistics?

Descriptive statistics organise observed data without generalising beyond the analysed sample. They answer questions such as: How many participants contributed data? What was the typical anxiety score? How variable were reaction times? What proportion completed each condition? Were scores symmetric, bounded, clustered, or strongly skewed?

Inferential statistics address uncertainty about population parameters or model comparisons. The distinction matters. A sample mean is a description. A confidence interval around an estimated population mean is inferential. A percentage describes the analysed participants, while a model may estimate a population probability after accounting for sampling and design.

APA’s Journal Article Reporting Standards for quantitative research expect clear participant accounting, data diagnostics, and results sufficient to understand the analysis. Descriptive statistics support each requirement, but only when their denominator, variable definition, and analytic sample are explicit.

Match the summary to the variable

Begin with the measurement level, distribution, and research purpose. A participant’s age in years, an ordered satisfaction category, a diagnostic group, and a reaction time are different data types. Forcing all of them into a mean and standard deviation produces numbers that may be mathematically possible but substantively unhelpful.

Do not choose solely from a normality test. Inspect the measurement process, the observed plot, sample size, outliers, and the analysis model. The NIST e-Handbook emphasises exploratory data analysis as a combination of summaries and graphics rather than a single numerical rule.

Mean, median, and mode

When the mean is informative

The arithmetic mean uses every value and represents the balancing point of a distribution. It is useful for interval-like or ratio variables when the scale and distribution make an average meaningful. The mean pairs naturally with the standard deviation for approximately symmetric data.

Because every observation contributes, extreme values can move the mean. That sensitivity is not automatically a defect. A very high valid symptom score is part of the sample and may legitimately affect the average. Investigate unusual values using the outlier-analysis guide rather than replacing the mean simply to suppress them.

When the median is informative

The median divides ordered observations in half. It is resistant to extreme values and often describes the centre of a skewed distribution better than the mean. Reaction times, waiting times, income measures, counts, and some symptom scores may be right-skewed, making the median and interquartile range useful.

Report enough context to prevent the median from hiding distribution shape. Two samples can share a median but differ greatly in spread, tails, or clustering. Include quartiles, sample size, range when relevant, and a plot.

When the mode helps

The mode is the most frequent observed value or category. It can describe nominal variables and reveal peaks in discrete distributions, but it is unstable in small samples and may not be unique. It rarely replaces a complete frequency table for demographic or categorical data.

Standard deviation, variance, range, and IQR

The standard deviation describes dispersion around the mean in the variable’s original units. NIST defines it as a measure based on squared deviations and notes its efficiency for normally distributed data. A small standard deviation means scores cluster relatively closely around the mean; a large one indicates greater dispersion.

Do not confuse standard deviation with standard error. The standard deviation describes variation among observations. The standard error describes the uncertainty of an estimator under a sampling model. CONSORT guidance explicitly warns that standard errors and confidence intervals are not substitutes for describing participant variability.

Variance is the squared standard deviation and is central to modelling, but its squared units are less intuitive in a descriptive table. The range gives the observed minimum and maximum, making impossible values and scale coverage visible, yet it depends heavily on sample size and extremes.

The interquartile range spans the 25th to 75th percentiles and describes the middle half of ordered values. Report it as Q1 to Q3 or as a width, following a consistent convention. “Median 12, IQR 8 to 17” is clearer than an unlabeled “12 (9).”

Counts, percentages, and denominators

Categorical variables require counts and proportions. Always state the denominator, especially when data are missing or participants could select multiple responses. A statement that “30% reported counselling” is ambiguous unless readers know whether the denominator is all recruited participants, all respondents to that item, or all participants retained in the analysis.

Report mutually exclusive categories that total the stated denominator. For multiple-response items, explain that percentages can sum above 100%. Avoid reporting percentages without counts in small samples because a seemingly precise percentage may represent only one or two people.

Use inclusive, globally appropriate demographic categories and explain self-description procedures. Collapse categories only when ethically and analytically justified, and do not combine small groups merely to make a table look tidy. Protect confidentiality where rare combinations could identify a participant.

Describe distributions, not only columns

Skewness and kurtosis coefficients can supplement a plot, but their definitions vary across software. NIST documents multiple skewness formulas and notes that some programs report excess kurtosis while others report kurtosis. Name the statistic and software convention if these numbers affect a decision.

A histogram shows shape and potential multimodality. A box plot highlights quartiles and possible extremes. A scatterplot reveals bivariate form, clusters, and influential observations. A trajectory plot shows individual change. Select a display that corresponds to the analysis rather than automatically exporting every chart.

A bimodal score distribution can signal subgroups, condition mixing, or a measurement process that a single mean cannot represent. Similarly, a narrow mean difference may conceal participants changing in opposite directions. Descriptive exploration helps you identify these features before formal modelling.

Descriptive statistics for psychology scale scores

Before summarising a multi-item scale, confirm scoring rules, reverse-coded items, allowable missing items, and the interpretation of total or mean scores. Use the instrument manual or validation study, not an improvised rule. The reliability and validity guide explains why scale consistency and measurement evidence remain separate from distribution summaries.

Report the theoretical range and observed range where floor or ceiling effects matter. A wellbeing mean of 24 is difficult to interpret without knowing whether the scale runs from 0 to 30 or 0 to 100. If many participants score at a boundary, show the proportion because central tendency and standard deviation alone may hide limited sensitivity.

Item-level Likert responses are ordered categories. A multi-item score created under an established scoring model may be treated as approximately continuous, but explain the convention and support it with the instrument literature. Do not claim that a mean proves equal psychological distance between response options.

Missing data changes every descriptive statistic

Each summary should make its valid sample size clear. If age has 200 observations but an anxiety score has 184, a table without variable-specific n values can mislead readers into assuming a common denominator. Report missing counts or percentages and describe how scale scores were calculated when some items were missing.

Do not replace missing values with zero unless zero is the genuine observed value. Do not silently use different samples across tables and models. The missing-data guide covers mechanisms, complete-case analysis, imputation, and sensitivity checks.

Describe the sample and participant flow

A sample table normally includes characteristics relevant to external validity, interpretation, and prespecified analyses. These may include age, gender, location, education, clinical status, language, or baseline measure, depending on the study. Avoid collecting or reporting attributes without a research or ethical rationale.

Separate recruitment flow from analysed characteristics. State numbers approached, screened, eligible, consented, allocated, completing each stage, and included in each primary analysis. These counts make attrition and exclusions visible. Link them to the sampling strategy and approved eligibility criteria.

Randomised-group baseline tables

For randomised designs, describe baseline characteristics by assigned group. CONSORT recommends means and standard deviations for suitable continuous variables, medians and percentiles for asymmetric data, and counts and proportions for categories. It advises against significance tests of baseline differences because randomisation already explains chance differences.

Do not use a baseline p value to decide whether to adjust for a covariate. Covariate adjustment should follow the design, theory, and prespecified analysis plan. In nonrandomised work, descriptive imbalance can be important but a table alone does not remove confounding.

Within-group, between-group, and time-specific summaries

Summarise variables at the level where the research question operates. An experimental study should show outcome summaries by condition. A longitudinal study should show each occasion and participant retention. A multilevel study may need participant-level and cluster-level summaries. A pooled mean can conceal important structure.

A defensible descriptive-statistics workflow

Step 1: Build a variable dictionary

List variable name, label, role, unit, permissible range, missing code, scoring rule, and measurement level. This prevents a numeric category code from being averaged or a sentinel missing value from being treated as a real score.

Step 2: Verify data before summarising

Check duplicate IDs, impossible values, units, condition codes, timestamps, and instrument ranges. Preserve raw data and run cleaning with syntax or a logged workflow. See the SPSS psychology dissertation workflow for reproducible setup.

Step 3: Plot every primary variable

Inspect overall and group-specific distributions. For repeated data, inspect trajectories and change scores. Record unexpected skewness, clusters, boundaries, and influential observations, then investigate them before inferential analysis.

Step 4: Select summaries based on meaning

Use mean and standard deviation when they provide a meaningful account of centre and spread. Use median and quartiles for skewed or resistant description. Use counts and proportions for categories. Report more than one representation when each answers a useful question.

Step 5: Make denominators visible

Report total recruited, total analysed, group n values, and valid n by variable when these differ. Include missingness. Avoid excessive decimal places that imply more measurement precision than the instrument provides.

Step 6: Create reader-ready tables

Build tables in the document rather than pasting raw software output. Give an informative title, define abbreviations in notes, include units, and state whether entries are mean and standard deviation, median and IQR, or n and percentage.

Step 7: Cross-check tables against analysis data

Verify that the sample in the descriptive table matches the corresponding model or clearly explain any difference. Recalculate totals and percentages, confirm group labels, and compare values with analysis output. Version-controlled syntax reduces transcription errors.

How to present descriptive statistics in APA style

Use prose for the few numbers essential to the narrative and tables for larger sets. Do not repeat every table value in text. Refer to the table, highlight the pattern that matters, and reserve interpretation of inferential findings for the appropriate section.

Italicise statistical symbols according to current APA conventions, define abbreviations, and use consistent decimal places. A conventional sentence might read: “Participants reported a mean stress score of 18.42 (SD = 5.31), with observed scores ranging from 6 to 30.” For skewed data: “Completion time had a median of 11.4 minutes (IQR 8.6 to 16.2).”

Exact style depends on institutional requirements, so follow your programme template where it differs. The results-section guide explains how to integrate descriptive and inferential findings.

Psychology-specific examples

Online mental-health survey

A student surveys stress, social support, and sleep among postgraduate students. The sample table reports age with mean and standard deviation, gender and study mode with counts and percentages, and country region with privacy-preserving categories. Variable-specific n values show that sleep has more missing responses than stress.

Stress is approximately symmetric, so mean and standard deviation are reported. Sleep latency is right-skewed, so the median and IQR are reported with a histogram. A scatterplot of stress and support precedes the correlation analysis, revealing a nonlinear cluster that requires investigation.

Repeated-measures attention task

A student compares neutral and emotional trials within participants. The results table reports valid participants, valid trials per condition, median trial response time, participant-level mean response time, accuracy, and within-person difference. This separates trial variability from participant variability.

Descriptive plots show that a small group traded speed for accuracy. Reporting response time alone would have hidden that pattern. The student retains both outcomes and aligns the inferential model with the preregistered analysis.

Common mistakes

• Using mean and standard deviation for every variable without checking meaning or shape.
• Reporting standard errors as if they describe participant variability.
• Giving percentages without counts or denominators.
• Averaging category codes such as gender, diagnosis, or region.
• Hiding missing data by omitting variable-specific n values.
• Reporting only pooled descriptives for grouped or repeated designs.
• Copying unedited software tables into the dissertation.
• Using excessive decimal places.
• Running baseline significance tests after random assignment.
• Calling a distribution normal based only on one test.
• Repeating every table number in prose.

Descriptive-statistics checklist

Frequently asked questions

Should I report mean and median?

Report the summary that best represents the variable and question. Reporting both can be helpful when skewness matters, but pair them with spread and a plot rather than presenting extra numbers without purpose.

What is the difference between SD and standard error?

The standard deviation describes variability among observed values. The standard error describes uncertainty in an estimator under a sampling model. Use SD, IQR, or category proportions to describe participants.

How many decimal places should I use?

Use enough precision to represent the measurement and analysis, not the software’s maximum output. Keep comparable values consistent and avoid implying accuracy the instrument cannot support.

Do I need descriptive statistics for every variable?

Describe variables needed to understand the sample, data quality, and analyses. Administrative identifiers and intermediate computation fields do not belong in the results table.

Can descriptive statistics prove a hypothesis?

No. They show the observed sample pattern. Population inference requires an appropriate design and inferential framework. Descriptive magnitude can still be substantively important and should be interpreted alongside effect estimates.

Conclusion

High-quality psychology dissertation descriptive statistics make the data visible before modelling begins. Match each summary to the variable, show centre and spread together, plot distributions, report valid denominators, and preserve the structure of groups, occasions, trials, and clusters.

Transparent description strengthens error detection, model choice, reproducibility, and interpretation. If you need ethical academic support, seek guidance that helps you understand and construct your own tables while keeping all analysis decisions, data handling, and authorship entirely yours.

Authoritative references

• American Psychological Association: JARS-Quant
• NIST/SEMATECH e-Handbook of Statistical Methods
• NIST statistical glossary
• NIST measures of skewness and kurtosis
• Guidelines for statistical reporting in research
• EQUATOR Network: SAMPL statistical reporting guidance
• CONSORT baseline-data guidance