Descriptive vs Inferential Statistics: A Plain-English Guide

Descriptive and inferential statistics are often introduced together in research methods courses, which can leave the actual distinction between them feeling blurry. In practice, they do genuinely different jobs, and most dissertations involving quantitative data need both - just not for the same purpose. This guide sets out what each one actually does, in plain English, with examples.

Descriptive Statistics: Summarising What You Found

Descriptive statistics do exactly what the name suggests: they describe and summarise the data you have collected, without making any claims that go beyond that specific dataset. They answer the question "what does my data actually look like?"

Common descriptive statistics include measures of central tendency (mean, median, mode), measures of spread (range, standard deviation, variance), and simple frequency counts and percentages. If you report that the average age of your survey participants was 34, that 62% of respondents identified as female, or that test scores ranged from 45 to 98, you are using descriptive statistics.

Descriptive statistics make no claim about anything beyond the specific sample you actually measured. Reporting that your sample's average score was 72 says nothing, on its own, about what the average score would be in a different sample or in the wider population - it simply describes what you found in the data you collected.

Inferential Statistics: Drawing Conclusions Beyond Your Sample

Inferential statistics go a step further: they use your sample data to make claims, with a stated level of confidence, about a larger population that you did not directly measure. They answer the question "based on what I found in my sample, what can I reasonably conclude about the broader population or relationship I am actually interested in?"

Common inferential statistics include t-tests, ANOVA, chi-square tests, regression analysis, and correlation coefficients with associated significance testing. If you report that a t-test showed a statistically significant difference between two groups, or that a regression model found income significantly predicts life satisfaction, you are using inferential statistics.

The key feature of inferential statistics is that they involve inference - a logical leap, made carefully and with a quantified level of uncertainty (typically expressed as a p-value or confidence interval), from your specific sample to a broader claim about the population your sample was drawn from.

A Concrete Worked Example

Imagine you survey 200 university students about study habits and exam performance.

Descriptive statistics would tell you: the average number of hours studied per week was 12.4, with a standard deviation of 4.1; 58% of students reported studying primarily in the evening; the average exam score was 74%.

Inferential statistics would tell you: a correlation analysis found a statistically significant positive relationship between hours studied and exam score (r = 0.42, p < .001), suggesting that this relationship likely holds true in the broader population of university students beyond just your specific sample of 200.

Notice the difference precisely: the descriptive statistics simply report what was true of your 200 respondents. The inferential statistic makes a claim - with a stated statistical confidence - about a relationship that likely exists more broadly, beyond just the people you happened to survey.

Why Most Dissertations Need Both

Almost every dissertation involving quantitative data needs descriptive statistics first, simply to establish and communicate what your sample actually looked like - this is standard practice regardless of your specific research question, and typically appears early in your results chapter (often called something like "sample characteristics" or "descriptive findings").

Whether you also need inferential statistics depends entirely on your research question. If your research question is purely descriptive - for example, "what proportion of nursing students report experiencing burnout?" - descriptive statistics alone may be sufficient to answer it, since you are not trying to test a relationship or make a broader causal or comparative claim. If your research question involves comparing groups, testing a relationship, or making a claim intended to generalise beyond your specific sample - for example, "does a specific intervention reduce burnout more effectively than standard practice?" - you need inferential statistics to properly test that claim, since a purely descriptive comparison of two averages cannot tell you whether an observed difference is a real effect or simply due to chance in your specific sample.

A Common Mistake: Treating Descriptive Findings as if They Were Inferential

A frequent error in student dissertations is describing a difference between two groups in purely descriptive terms - "Group A scored higher on average than Group B" - without running the appropriate inferential test to establish whether that difference is statistically meaningful or could plausibly have occurred by chance. A difference in sample averages that has not been tested inferentially cannot be presented as a genuine finding about a real underlying difference between groups; it is simply a description of what happened to be true in your specific sample, which could easily look different in another sample of the same size purely due to random variation.

The reverse mistake also happens: running an inferential test but reporting only the p-value ("p < .05, so the result is significant") without also reporting the actual descriptive statistics (the means, the effect size) that give the result any real-world meaning. A statistically significant result with almost no practical difference between groups is not necessarily an important or meaningful finding, and readers need the descriptive numbers alongside the inferential test result to judge that for themselves.

Choosing the Right Inferential Test

Selecting the correct inferential statistical test depends on several factors: the type of data you have (categorical, ordinal, or continuous), how many groups you are comparing, whether your data meets the assumptions required for parametric tests (such as normal distribution), and precisely what kind of relationship or difference you are trying to test. Getting this wrong - for example, running a test designed for normally distributed data on data that clearly is not - can produce a genuinely misleading result even if every other part of your methodology is sound. If you are unsure which test fits your specific research question and data, this is worth confirming with your supervisor or a statistics consultant before running your analysis, rather than after.

Getting Support With Your Statistics Chapter

Choosing the right combination of descriptive and inferential statistics, and presenting both clearly and accurately, is one of the areas where dissertation students most often want a second opinion before submission. If you want feedback on whether your statistics section presents both types clearly and uses inferential tests appropriately for your actual research question, CampusScribe's editors review dissertation methodology and results chapters for exactly this kind of statistical clarity and accuracy.

From choosing a topic to polishing a final draft, CampusScribe's subject-specialist editors and coaches help with dissertations, capstone projects, presentations, and more - whether you need topic guidance, structural feedback, or a final edit before your deadline.