Under Grade test → Statistics, you see how your class did: the grade distribution, the performance per label and, for each question, figures like the p-value and Rir-value. Click Generate summary and the AI sums up what stands out. On the Teacher and Organization plans, it also suggests groups of students to follow up with.
It's the last step of creating, delivering and grading tests with AI: you see whether the test was too hard, which question didn't work, and where your class struggled.
Where to find it
Open the test and go to Grade test → Statistics. From top to bottom, you'll find the Summary, the Grades distribution, Performance per label, Suggested groups and the Statistical analysis.
Examplary includes AI suggestions you haven't reviewed yet in the numbers. While anything is still open, you'll see Not all answers have been graded yet. Some statistics may be inaccurate.
Getting a summary
Click Generate summary. The AI reads your class's statistics, without names or answers, and writes a few sentences on how it went. Below that are the Main concern and what's Also notable, with links to the questions they're about.
The summary is saved with the test, so your colleagues see the same one. If the results change, you'll see Update summary. It's still the AI's reading of the numbers: check it before you act on it.
Reading the grade distribution and performance per label
Under Grades distribution, you see the lowest, highest and average score, the standard deviation and a chart with the number of students per grade. A low standard deviation means most students scored close to the average. How points turn into a grade is set in the grade calculation.
Under Performance per label, you see for each label, such as a level from your taxonomy, what percentage of answers was correct. Only answers with full points count as correct. See Using taxonomies in Examplary.
Suggested groups
After the summary, the AI suggests up to three groups: Reteach, Practice and Stretch. Each group comes with a short suggestion and the names of the students in it. Click a name to see that student's results.
To give a group more practice, make a practice space from the test and invite those students. See Turning a test into a practice space.
Suggested groups are available on the Teacher and Organization plans. The summary and the other statistics are available to everyone.
Reading the statistical analysis per question
This table shows how well each question worked. Hover over a column name for an explanation.
| Column | What it tells you | Rule of thumb in the app |
|---|---|---|
| Graded answers | How many answers have been graded | |
| p-value | How easy the question was: the average share of the points students got, from 0 to 1. Partial credit counts. | Above 0.90 is very easy, below 0.10 may be too hard |
| Rir-value | Whether students who did well on this question also did well on the rest of the test | Above 0.3 is good, below 0.2 is a reason to look at the question again |
| Rit-value | The same, but compared with the whole test, this question included | As for the Rir-value |
| p'-value | The difficulty corrected for guessing, only for questions with answer options, like multiple choice |
If a value can't be calculated, for example because there are too few answers, you'll see a dash.
Taking a question out of the grading
If a question didn't work, for example because of a mistake in it, take it out of the grading. Go to the Questions view, open the question and choose Exclude from grading in the three-dot menu. Examplary recalculates everyone's points and grades. To undo it, choose Include in grading.
If you've already published the results, students see the new grade straight away.
Exporting results
With the export icon, you can make a Print PDF with an overview and a page per student, or choose Download spreadsheet for each student's grades, points, answers and feedback. The statistics and the summary aren't part of the export. Students never see this overview.
Troubleshooting
It says 'Could not generate the summary'. No student has any points yet, or something went wrong. Grade a few answers first and try again.
There are dashes in the table. There are too few graded answers, or every student scored the same on that question. The p'-value only exists for questions with answer options, like multiple choice or true/false.
Suggested groups tells me to upgrade. They're not included in your plan.