Upload a CSV or Excel file, select a numeric column, and get a full Minitab-style analysis: descriptive statistics, normality test (Anderson-Darling), histogram, box plot, Q-Q plot — with specific guidance if your data is non-normal.
Download one of the sample files below, then drag it into the upload zone above — or click the zone to browse for it. Upload one file at a time and watch the full analysis unfold. Start with the skewed dataset to see how non-normal data looks, then try the normal one to compare the difference.
Use the calculator above to upload a CSV or Excel file, select a numeric column and get a full Minitab-style analysis: descriptive statistics, an Anderson-Darling normality test, histogram, box plot and Q-Q plot — with specific guidance if your data is non-normal. It is a complete first-pass diagnostic before any deeper Six Sigma analysis.
A Dataset Analysis is the structured exploratory phase of any statistical study. It combines descriptive statistics (mean, median, σ, skewness, kurtosis), a formal normality test, and visualisations (histogram, box plot, Q-Q plot) to characterise the data before any modelling, capability calculation or hypothesis test is run.
The tool computes the standard descriptive statistics, runs the Anderson-Darling test for normality (more powerful than Shapiro-Wilk for tails), and generates three plots. If the data fails the normality test, it provides specific guidance — try a Box-Cox transformation, switch to non-parametric tests, or use non-normal capability methods.
You upload 500 cycle-time measurements. The tool reports mean 12.3s, median 11.8s, σ 3.5s, skewness +1.4 (right-skewed), Anderson-Darling p < 0.005 (clearly non-normal). The histogram confirms a right tail; the Q-Q plot shows the upper end pulled away from the line.
The tool flags that running Cpk or a t-test on this data without transformation would mislead. After a log transform the data is approximately normal, and the analyses become valid. That single diagnostic step prevents weeks of wrong conclusions in DMAIC Analyse.
Dataset Analysis catches data-quality and distribution issues before they corrupt downstream analysis. It is the difference between a Six Sigma project built on a defensible foundation and one quietly built on invalid assumptions.
Use it on every dataset before computing capability indices, running hypothesis tests, building regression models or designing experiments. It is the first step in DMAIC Measure and Analyse.
Dataset Analysis sits between data collection and inferential analysis. It pairs with Measurement System Analysis (MSA) to confirm data is both trustworthy and statistically tractable before any decisions are made.
If the data is normal, proceed with standard capability, hypothesis tests and DOE. If non-normal, transform (Box-Cox or log), switch to non-parametric methods, or use non-normal capability techniques. Either way, document the choice in the project file.
Pair Dataset Analysis with Measurement System Analysis, control charts and capability indices for a complete pre-analysis workflow.
A summary of a data set using measures of central tendency (mean, median), spread (σ, IQR, range) and shape (skewness, kurtosis). It is the starting point of any analysis.
A formal hypothesis test for normality, more sensitive in the tails than the Shapiro-Wilk test. A small p-value (typically < 0.05) indicates the data is not normally distributed.
Three options: (1) transform the data (Box-Cox, log), (2) switch to non-parametric tests (Mann-Whitney, Kruskal-Wallis), or (3) use non-normal capability methods. The tool flags the recommended path.
A Q-Q plot reveals where data deviates from a reference distribution. A straight line means normal; bends in the tails reveal skew or heavy tails. It is more diagnostic than any single test statistic.
Not always. If the analysis you plan is robust to non-normality (e.g. ANOVA on large samples), transformation may be unnecessary. The right answer is to match the method to the data, not force the data to fit the method.
Want to learn how to structure data analysis as part of a full improvement project? The Green Belt covers this in full.
View Green Belt →