Analysis problems

Calculate results saves the study, then calculates from the saved data. If it can't produce a trustworthy result, it stops and says why. Your entries remain available for correction. See also Calculate results.

What you see. Fix the marked rows before analysis. (in some study types, Resolve the marked rows before analysis.), and Data opens with Fix errors on each problem row. A count above the grid says how many entries need checking. Go to first entry takes you to the first affected cell; Show details lists the problems, for example Row 3 · Result.

Why. An included row has a blank result, text such as <0.5 or hemolyzed, a number in the other numeric convention, or a missing ID.

Recover.

1. Choose Go to first entry, or open Show details and select a cell. Enter the missing measurement or correct the mistyped value. The old error message clears as you edit.
2. For a result that must be left out, choose Fix errors on its row and exclude it with a reason. If the blank rows were not measured, the notice's Exclude n rows button excludes them together; completely empty rows are removed. See Blank results and Correct or exclude a result.
3. Choose Calculate results again to check the corrected entries and save the calculation.

Row messages are listed in Validation messages.

What you see. The main button reads Finish set up, and the line under the stage tabs names what is needed. See What stops a calculation.

Recover. Fill in the listed fields on Set up, for example the unit, an assigned value, claimed reference limits or the lots. Then calculate again.

What you see. A message about the design, often naming the level. For example:

• You need results from at least 2 days. (multi-day precision)
• At least one run needs 2 or more replicates, so that repeatability can be estimated. (multi-day precision; a lost replicate, a short run or a day with fewer runs is calculated, with a study note that names it)
• Enter paired results for at least two specimens before analysis. (method comparison)
• Enter at least one included result before analysis.

Why. The data don't have the structure the calculation needs. The chapter for your study type gives its design rules. See Study methods.

Recover.

• Add missing measurements, or exclude extras with a reason, to complete the design.
• Make day, run, replicate and level labels consistent. 1 and Day 1 are different days.
• Check exclusions. Excluded rows don't count toward minimum numbers.
• If your design is different, use the study type that fits it.

What you see. A banner says the results are out of date, and the button reads Recalculate results. Results may also say the acceptance limits changed after this calculation.

Why. Results come from the data as saved when they were calculated. Any later edit to rows, settings or limits makes them out of date.

Recover. Choose Recalculate results. If you are viewing an earlier calculation, choose Show latest calculation.

What you see. The study is Undecided and a limit is Undecided with a reason in the acceptance table, or a limit is Not applicable.

Why. Common reasons:

See About acceptance decisions and Set acceptance limits.

What you see after Save changes. Changes to different rows or settings are combined. If both people changed the same value, Check the combined changes opens and names the overlap. It also opens when the other save removed a row you had changed, for example by importing a file that replaced the current rows; your changed row is kept at the end.

Why. A colleague, or another of your tabs, saved the study after this page opened it.

What to do. Check the combined entries. If a kept row's result is already among the other person's rows, remove the kept row so it is not counted twice. Then choose Save changes in the dialog. Review first returns to the study with your draft intact. See When someone else saved first.

When calculating, signing off or importing. Calculating and importing save your edits first, so an overlap opens the same dialog. Check it and save. Without unsaved edits, or when signing off, a red banner says another version was saved. Reopen the current study and try again.

A save either completes or changes nothing.

• The saved study has values that need fixing. Open the study, fix and save them, then calculate again.
• A precision study linked for the Deming λ has changed since this study was calculated. … On Set up, in Method comparison, choose the current results or another basis for λ, then save and calculate again. See Method comparison. The message appears when you calculate again without changing anything after a linked precision study changed: the earlier calculation used the result that was current then, so it isn't returned again. A calculation after any other change still runs with the linked result and leaves the Deming limits Undecided until you choose a basis for λ.
• The spread is too small to calculate reliably. The results differ by amounts too small for the arithmetic. Check for extreme values or unit errors.