Descriptive Statistics
The first step in every analysis — means, standard deviations, skewness, kurtosis, and a correlation matrix, ready to report in APA.
What it is
Descriptive statistics summarise your data before any modelling begins. For each continuous variable you get the mean (M), standard deviation (SD), minimum, maximum, and range. Skewness and kurtosis show whether the distribution is symmetric and whether the tails are heavy — useful checks before running parametric tests. For categorical items, frequency tables show how responses are spread across categories. A Pearson correlation matrix shows the bivariate relationship between all continuous variables at once. Explonia also counts missing data per variable so you can see where responses are incomplete.
When to use it
- You have just collected or uploaded data and need a first look at your sample.
- You want to check item distributions and spot unusual values before running a model.
- You need to report sample characteristics (means and SDs) in a table.
- You want to inspect bivariate correlations before running regression or SEM.
- A reviewer or supervisor asks for a correlation matrix.
When not to use it
- Descriptives do not test differences or relationships — they describe. Run a t-test, ANOVA, or regression for inferential conclusions.
- A correlation does not imply causation and does not control for other variables — use regression or SEM for that.
Assumptions
- Continuous variables for the mean, SD, skewness and kurtosis; categorical variables for frequency tables.
- Skewness and kurtosis are read as rough checks before parametric tests, not as hard pass/fail rules.
- Pearson correlations assume roughly linear relationships between continuous variables; inspect for non-linearity and outliers.
- Missing data is reported per variable; more than 5% missing on an item is worth examining.
Thresholds & cutoffs
| Criterion | Rule of thumb | Source |
|---|---|---|
| Skewness | |skew| < 2 is often cited as acceptable for parametric tests | George & Mallery (2010) |
| Kurtosis | |kurt| < 7 is often cited as acceptable | West, Finch & Curran (1995) |
| Correlation (r) | Sign shows direction; |r| ≥ .10 small, ≥ .30 moderate, ≥ .50 large (rough guides) | Cohen (1988) |
| Missing data | Flag variables with > 5% missing; report counts per item | Tabachnick & Fidell (2013) |
How to report it in APA
Participants (N = 248) showed moderate satisfaction (M = 4.21, SD = 0.87, range 1–7). The correlation matrix (Table 1) showed that price sensitivity and brand loyalty were negatively associated (r = −.34, p < .001). Skewness and kurtosis values were within acceptable ranges (|skew| ≤ 1.3, |kurt| ≤ 2.1), supporting the use of parametric analyses.
How Explonia runs it
Open any study in Explonia and choose Descriptives under Analyses. Explonia computes per-variable statistics — M, SD, min, max, skewness, kurtosis, and missing count — plus a full Pearson correlation matrix with significance flags. The Article description box drafts an APA paragraph from your numbers that you can edit before copying. All tiers include descriptive statistics; the only limit across tiers is the number of responses in your study.
Frequently asked questions
What do skewness and kurtosis tell me?
Skewness measures how asymmetric the distribution is; kurtosis measures how heavy the tails are. Values of |skew| below 2 and |kurtosis| below 7 are commonly cited as acceptable for parametric tests, though these are rough guides, not hard rules. Moderate violations are less problematic with larger samples.
Should I report a full correlation matrix in my paper?
For survey and scale data, a correlation matrix shows how variables relate to each other before modelling. Many journals and reviewers expect it. Explonia generates the matrix automatically and drafts the APA paragraph for you.
How many decimal places does APA require for descriptive statistics?
APA 7 recommends two decimal places for means, standard deviations, and correlations. Use the same number of decimal places consistently within a table.
How much missing data is acceptable?
There is no universal cutoff, but missing data above 5% per variable is worth examining. Report the count or percentage of missing values per item. If missing data is not random (MCAR), imputation or other methods may be needed — check with your supervisor or journal guidelines.
Last updated: July 2026
Run Descriptive Statistics on your own data
Collect or upload your data, run the analysis in your browser, and get a draft APA paragraph you can edit. Free to start. No credit card.
Start free