Correlation analysis

Correlation analysis without writing a script

A correlation coefficient is one number, and most of the real work is choosing the right coefficient and reading the interval around it. Magic Stat gives you Pearson, Spearman and Kendall with their confidence intervals, plus partial correlation — from the spreadsheet you already have.

The honest part

What a correlation coefficient does and does not say

A correlation coefficient summarises how strongly two variables move together. Pearson's r measures the strength of a linear relationship between two continuous variables. Spearman's rho and Kendall's tau work from the ranks instead, so they describe any monotonic relationship and are not thrown off by skew, outliers or ordinal data. That is the first decision: the coefficient has to match the shape of the data, not the other way round.

The second thing a coefficient does not say is causation — and it also does not say how precise the estimate is. Magic Stat reports the confidence interval next to every coefficient, because an r computed on a handful of pairs and the same r computed on hundreds are not the same result.

When two variables look associated only because both move with a third, the useful analysis is partial correlation: the correlation between X and Y after the linear effect of the covariates is removed. Magic Stat does that in the same dialog.

What it does

What Magic Stat gives you for correlation

How it works

Two columns in, one interval out

  1. Load the spreadsheet. each variable in a column — no reformatting, no transposing.
  2. Pick the coefficient. Pearson for roughly linear continuous pairs; Spearman or Kendall for monotonic, skewed or ordinal data.
  3. Select X and Y. and, for partial correlation, the covariates to hold constant.
  4. Read the interval, not just r. the coefficient, its confidence interval, n and the p-value are shown together.
  5. Export. the analysis and its interpretation go into the report as .docx, .html or .md.
Options

The choices, out in the open

Coefficient: Pearson · Spearman · Kendall.

Interval: Fisher z for Pearson, Bonett–Wright for Spearman, bootstrap for Kendall — matched automatically to the coefficient you chose.

Partial correlation: choose the covariates to control for; the result is the correlation of the residuals.

Frequently asked

Correlation questions, answered honestly

Which coefficient should I use?

Pearson if both variables are continuous and the relationship is roughly linear. Spearman if the relationship is monotonic but not linear, or the data are skewed or ordinal. Kendall if there are many ties or the sample is small and you want a coefficient defined in terms of concordant pairs.

What value counts as a strong correlation?

There is no universal threshold, and treating one as if there were is a common mistake. What matters is the coefficient together with its confidence interval and the sample size. Magic Stat labels the strength for convenience, but the interval is the honest part.

Does correlation imply causation?

No, and no software can make it. If you suspect a third variable drives both, use partial correlation to hold it constant — that still does not establish a causal direction.

Do I need R?

No. If you already work in R and it serves you, keep it — R's cor.test is the reference this dialog is built to match. The value here is the interval, the partial correlation and the report in one place.

Is my data uploaded somewhere?

No. The analysis runs on your computer. The only signal the app sends automatically is an anonymous installation counter — no data, no names, no e-mail.

Linear regression → Robust regression → Descriptive statistics →