Detrending FAQ · analysis without uploadFrequently asked questions
This detrending FAQ covers trend testing, method selection and analysis without upload. If a question is missing, the method notes go deeper.
Common questions about detrending safety, Mann-Kendall method selection, multi-sheet Excel compatibility and data privacy.
Detrending FAQ: analysis without upload
Does my Excel file get uploaded anywhere?
No. The file is read with the browser File API, parsed inside a Web Worker, and analysed in a second worker on the same device. There is no upload endpoint, no server action and no API route that accepts file bytes. Closing the tab discards the data.
Which file formats and sizes are supported?
XLSX, XLS and CSV up to 15 MB. Multi-sheet workbooks are supported: sheet names are read first, the sheet with the most numeric columns is selected automatically, and other sheets are materialised on demand.
Why does the tool refuse to detrend some series?
Because a straight line through a random walk invents structure. When ADF cannot reject a unit root and KPSS rejects stationarity, deterministic detrending is marked as blocked. You can still override it, but the result carries a permanent warning and the differencing alternative is shown.
What does a p-value mean here?
It is the chance of seeing a trend at least this strong if the series had no trend at all. A p-value of 0.003 means roughly 3 in 1,000 random arrangements would look this strong. It is not the probability that the trend is true, and it says nothing about effect size.
Why do different Mann-Kendall variants disagree?
They make different assumptions about independence and seasonality. TFPW removes autocorrelation and can weaken a real trend. Yue–Wang corrects the variance and can turn a significant result into a non-significant one. The workspace shows the corrected and uncorrected results side by side rather than choosing for you.
What is the difference between tau and the Theil–Sen slope?
Tau is the strength of the monotone relationship, scaled from −1 to 1 and insensitive to magnitude. The Theil–Sen slope is the median pairwise rate of change in the original units. A trend can be perfectly consistent (τ near ±1) while the rate of change is tiny.
How are outliers handled?
They are labelled, not silently deleted. The denoising pipeline records every removed row, the rule and the threshold that removed it, and the conclusion is checked with and without those points. If the outliers are genuine events, you can turn denoising off and compare.
What happens with missing values?
Missing values are excluded from the analysed series and the missing ratio is reported. Above 30% missing, the stationarity verdict becomes “insufficient evidence” and trend inference is blocked. The tool does not impute silently.
Can I analyse more than one column at once?
Yes. “Analyze everything” tests every numeric column and applies Benjamini–Hochberg FDR by default. The table reports significance before and after correction. Columns that lose significance after correction are not labelled significant.
Why are some methods marked UNVALIDATED?
Because no independent oracle was available for that specific correction. The implementation follows the published formula, but without a third-party reference it cannot enter the validated matrix. Those methods are shown with a permanent badge and never carry the headline conclusion when a validated method exists.
What is an independent oracle?
A separate implementation outside this project — statsmodels, SciPy, pymannkendall, or a published reference vector — whose output is compared against this engine on a deterministic fixture. A tolerance is set before the comparison and published with its rationale. You can re-run the whole matrix from the repository.
Does the tool work offline?
After the page has loaded, yes — as far as the analysis goes. All computation is local: the workbook is parsed in this browser, the statistics run here, and the cross-sheet lookup reads the worksheets already in memory. No database engine is loaded. Advertising slots, if present, are the one part that may still request content from a third-party network; they never receive your data.
Related method notes
Or open the workspace and run the analysis on your own workbook.
This page publishes FAQPage structured data so that each answer can be quoted on its own by search engines and answer engines. Google restricts FAQ rich results to authoritative government and health sites, so no rich result is claimed here; the markup is for machine reading, not for a star in the snippet.