Descriptive statistics

Descriptive statistics and a table ready for the paper

Every paper opens with a table of the sample. Magic Stat produces it: mean (SD), median (IQR), frequencies, normality and outliers per variable, and a descriptive table — with optional group columns and tests — that goes into Word without retyping.

The honest part

Describing the data, and the table you actually need

The descriptive module reports, for each numeric variable, the mean and standard deviation, the median and interquartile range, the minimum and maximum, skewness and kurtosis, the result of a normality test and the number of flagged outliers; for categorical variables it gives the frequency table. Histograms with a fitted distribution come with it.

The separate Descriptive Table for Paper turns those numbers into the table that opens a manuscript: n, mean (SD), median (IQR), range, skewness, kurtosis, normality and a 95% confidence interval, for the variables you select, optionally split by a grouping variable with the group test alongside.

This is the least glamorous part of statistics, and the honest position is that almost any software does descriptive statistics. The reason to use Magic Stat here is the table: it comes out formatted for the paper, with the grouping and the tests already in place, instead of being assembled by hand in Word. If descriptives are all you need, a spreadsheet is a fair choice.

What it does

What Magic Stat gives you for descriptives

How it works

From raw columns to the opening table

  1. Load the data. the variables you want described; numeric and categorical columns are treated accordingly.
  2. Choose the variables. for the explorer, or for the paper table — you select the subset to describe.
  3. Set the statistics and the normality test. n, mean (SD), median (IQR), range, skewness, kurtosis, normality and 95% CI, each toggled on or off.
  4. Add a grouping variable if you need one. the table then shows the statistics per group, with the test between them.
  5. Copy or export. paste the table into Word or Excel, or export the data to a file.
Options

The settings, in the dialog

Statistics to include: n · mean (SD) · median (IQR) · min–max · skewness · kurtosis · normality test · 95% CI. Normality method: Shapiro–Wilk (recommended, n ≤ 5000) · D'Agostino-Pearson (omnibus, n ≥ 20) · Kolmogorov–Smirnov with Lilliefors · Anderson–Darling.

Grouping: none (overall statistics) or a categorical column, with chi-squared or Fisher's exact for the categorical variables. Outliers in the explorer can be flagged by IQR, Z-score, MAD, Grubbs, Dixon's Q or Rosner ESD.

Frequently asked

Descriptive-statistics questions, answered honestly

Which normality test should I use?

Shapiro–Wilk for most samples up to a few thousand cases; D'Agostino-Pearson as an omnibus test for larger samples; Kolmogorov–Smirnov with Lilliefors when that is the convention in your field, and Anderson–Darling when you want more weight on the tails. Reporting more than one is defensible; picking the one that gives the answer you want is not.

Which outlier rule should I report?

Say which one you used, and report all of them if they disagree. IQR is the conventional rule; Grubbs and Rosner are formal tests; Dixon's Q only applies to small samples. The module computes them all so the choice stays visible rather than hidden.

Can the table be split by group, as in a Table 1?

Yes. Choose a grouping variable and the descriptive table shows each statistic per group, with a chi-squared or Fisher's exact test for the categorical variables.

Is it worth using this just for descriptives?

Honestly, only if the formatted table saves you real time. A spreadsheet will compute means and medians. The value here is the table that is ready for the manuscript and the outlier and normality battery in the same place.

Is my data uploaded somewhere?

No. The analysis runs locally; the only automatic signal is an anonymous installation counter with no data in it.

t-test → ANOVA → Correlation analysis →