Data Analysis Basics for Students: A Plain-English Guide
Many students approach data analysis with dread, picturing formulas and unfriendly software. The ideas behind most student-level analysis are simpler than they look, and a lot of the difficulty comes from not knowing which question to ask first. If you know what your data are, what you want to find out and what each common technique is for, you can make good choices and explain them.
This guide covers the main steps from raw data to reported results, with a short section on qualitative analysis. It is an introduction, so use your course materials and supervisor for details specific to your field.
Start with the question
Every analysis starts from your research question or hypothesis. Before you open any software, write down what you want to know and which variables are involved. Are you describing a group, comparing groups, looking for a relationship or predicting an outcome? That sets the direction. Our guide to writing a hypothesis explains how to state your predictions so that the analysis follows.
Know your data
Types of variables
| Type | Description | Example |
|---|---|---|
| Categorical (nominal) | Categories with no order | Course, gender, hospital unit |
| Ordinal | Ordered categories | Satisfaction: low, medium, high |
| Interval | Ordered with equal steps, no true zero | Temperature in Celsius, some test scores |
| Ratio | Equal steps and a true zero | Age, hours, income, counts |
Interval and ratio data are often called continuous or scale data. The level of measurement matters because it limits which statistics make sense. You can average hours studied, but averaging category codes for course is meaningless.
Independent and dependent variables
The independent variable is the factor you think has an effect, and the dependent variable is the outcome. Knowing which is which shapes both the test and the way you report.
Clean the data
Raw data are messy. Before analysis:
- Check for entry errors, such as impossible ages or values outside the scale.
- Look for duplicates.
- Handle missing data with a stated rule. Options include leaving cases out of specific tests, or, with guidance, using methods that estimate missing values.
- Recode variables where needed, for example reversing negatively worded items.
- Create new variables, such as total scale scores.
- Keep a log of every change, and keep a copy of the original file.
Decisions at this stage affect your results, so document them in your methodology.
Describe the data
Descriptive statistics summarize what you have. They come first, since they show you what the data look like.
| Statistic | What it tells you |
|---|---|
| Frequency and percentage | How many fall into each category |
| Mean | The average |
| Median | The middle value, less affected by extremes |
| Mode | The most common value |
| Range | The spread from lowest to highest |
| Standard deviation | How far values typically fall from the mean |
Use charts, such as histograms, bar charts and box plots, to look at distributions and spot outliers. Report descriptives for your sample, so readers know who you studied.
Choose an inferential test
Inferential statistics let you ask whether patterns in your sample are likely to hold in a wider population. The right test depends on your question and your data.
| Question | Variables | Common test |
|---|---|---|
| Do two groups differ? | One categorical (two groups), one continuous | Independent samples t-test |
| Did scores change in the same people? | Same participants measured twice | Paired samples t-test |
| Do three or more groups differ? | One categorical (3+ groups), one continuous | One-way ANOVA |
| Are two categorical variables associated? | Two categorical | Chi-square test |
| Are two continuous variables related? | Two continuous | Pearson correlation |
| Can one or more variables predict an outcome? | Predictors and a continuous outcome | Linear regression |
| Can variables predict a yes or no outcome? | Predictors and a binary outcome | Logistic regression |
When your data break the assumptions of these tests, such as normality, there are non-parametric alternatives, for example the Mann-Whitney U test or Spearman correlation. Ask your supervisor or a statistics tutor to confirm your choice before running many tests.
Check assumptions
Tests make assumptions about the data, such as roughly normal distributions, equal variances, independence of observations and adequate sample size. Check them, usually with plots and simple tests, and report what you found. If assumptions are violated, you can choose a different test or adjust. Ignoring assumptions is a common source of unreliable results.
Interpret the results
p-values
A p-value is the probability of getting results at least as extreme as yours if the null hypothesis were true. By convention, many fields treat a p-value below .05 as statistically significant, but this threshold is a convention, not a law. A small p-value suggests the result is hard to explain by chance, and a large one means you do not have enough evidence to reject the null. It does not tell you the size or importance of an effect, and it does not give the probability that your hypothesis is true.
Effect sizes
An effect size shows how big a difference or relationship is. Common ones include Cohen's d for differences between means, r for correlations and eta squared for ANOVA. Large samples can make tiny effects statistically significant, so report effect sizes to show whether a result matters in practice.
Confidence intervals
A confidence interval gives a range of plausible values for an effect. Reporting one shows the precision of your estimate. A wide interval means more uncertainty.
Report the results
Reporting usually follows a standard pattern. In APA style, for example, an independent samples t-test might be reported as:
Students in the spaced practice group (M = 78.4, SD = 9.1) scored higher than students in the block practice group (M = 72.9, SD = 10.3), t(58) = 2.17, p = .034, d = 0.56.
(The numbers are invented to show the format.) Include the test, the degrees of freedom, the test statistic, the p-value and an effect size, and explain the result in plain words. Use tables and figures for clarity, with numbered captions. Our guide to writing the results and discussion chapters shows how to organize and interpret findings.
Multiple tests
Running many tests increases the chance of a false positive. Decide your tests in advance, link them to your hypotheses, report all that you ran and consider adjusting for multiple comparisons if your field expects it. Avoid hunting through the data for anything with p below .05.
Common software
| Tool | Notes |
|---|---|
| Excel | Fine for cleaning, descriptives and simple charts, limited for advanced tests |
| SPSS | Point and click, widely used in social sciences and health |
| R | Free and powerful, with a learning curve |
| jamovi and JASP | Free, user-friendly and based on R |
| Python | Common in data science and computing |
| NVivo, ATLAS.ti, MAXQDA | Qualitative coding software |
Use the software your program supports, and keep a record of every step, such as saved syntax or scripts, so you can reproduce your analysis.
Qualitative analysis in brief
Qualitative analysis works differently. A widely used approach, thematic analysis, usually involves these steps:
- Become familiar with the data by reading transcripts repeatedly.
- Generate initial codes by labeling meaningful segments.
- Search for themes by grouping related codes.
- Review the themes against the data.
- Define and name the themes.
- Write up, using quotations to illustrate each theme.
Be transparent about how you coded, whether others checked your coding and how you resolved disagreements. Our guide to conducting research interviews explains how the data are collected.
Check your work
Before you report:
- Recheck data entry on a sample of cases.
- Rerun key tests to confirm results.
- Look at the plots along with the numbers.
- Check that numbers in text and tables match.
- Ask someone to review your approach, particularly your supervisor or a statistics help service.
A mini example from start to finish
Imagine a survey of 120 students asking about weekly study hours and exam scores. You clean the file and find three impossible values for hours, which you check against the original forms and correct. You describe the sample: mean study time, spread and the distribution of scores.
Your question is whether more study time goes with higher scores, so you plan a correlation, check that the relationship looks roughly linear on a scatter plot and run the test. You report the correlation, its p-value and a confidence interval, and explain in plain words what the strength of the relationship means. Finally, you note that correlation does not show that studying causes higher scores, because motivation or prior knowledge could affect both.
Getting help with statistics
Most universities offer free statistics support through a library, a learning center or a quantitative methods clinic. Bring your research question, your variables and a sample of your data, and ask which test fits. Short online tutorials from your software's maker are often the fastest way to learn a specific procedure. Whatever help you use, you remain responsible for understanding and explaining the analysis in your own words, since examiners may ask you to justify it.
Common mistakes
- Choosing a test before knowing the question.
- Skipping data cleaning.
- Ignoring assumptions.
- Reporting p-values without effect sizes.
- Treating "not significant" as "no effect."
- Claiming causation from correlation.
- Running many tests and reporting only the significant ones.
- Not documenting the analysis.
- Overinterpreting small samples.
A quick checklist
- Do I know my research question and variables?
- Have I cleaned and documented the data?
- Have I described the sample?
- Is my test matched to the question and data?
- Have I checked assumptions?
- Do I report effect sizes and interpret in plain words?
- Is my analysis reproducible?
Final thoughts
Data analysis is a series of small, sensible decisions: know your question, know your data, describe before you test, choose a test that fits and report honestly. Ask for help early, since a short conversation with a statistics tutor can save weeks. And remember that clear explanation of what you did and found matters as much as the numbers themselves.
Frequently asked questions
How do I choose a statistical test?
Match the test to your question, the type of variables and the design. Ask whether you are comparing groups, looking at relationships or predicting an outcome, how many groups you have and what kind of data each variable is.
What does a p-value mean?
It is the probability of seeing results at least as extreme as yours if the null hypothesis were true. A small value suggests the data are hard to explain by chance alone. It does not tell you how large or important an effect is.
What is an effect size?
A measure of how large a difference or relationship is, independent of sample size. Reporting it alongside the p-value shows whether a result matters in practice.
Which software should I use?
Whatever your program supports. Common choices include SPSS, R, Excel, and free tools such as jamovi and JASP. Ask your supervisor, and learn one tool well.
Do I have to clean my data?
Yes. Raw data usually contain errors, duplicates and missing values. Check and document how you handled them before analyzing, since it affects every result.