# Percentages for multiple survey selections or themes

Choose between respondent shares and selection shares when one person can contribute multiple survey categories.

If nothing was unclear or difficult, say so. Skip questions about steps you did not experience; use not applicable where appropriate.

## 1. Which issues, if any, did you encounter during the task?

Answer: ____________________

## 2. Which issue most affected your next step?

Answer: ____________________

## 3. Were two issues part of the same incident or separate experiences?

Answer: ____________________

## 4. What issue category is missing or poorly described?

Answer: ____________________

## Author notes / illustrative keys where applicable

1. Allow more than one issue and a none route without assuming difficulty.
2. Distinguish a primary concern from a list of all mentioned themes.
3. Avoid treating multiple mentions as independent people or incidents.
4. Check category meaning before coding free-text responses.

Interpretation caution: A multi-theme count is not a single-choice distribution. Define whether the unit is people, answers or mentions and do not imply statistical independence.


## Illustrative finding and follow-up

Observation: Synthetic coding from 12 people: connection issues mentioned by five, access by four and instructions by seven; 16 mentions include overlap.

Action to test: Report instructions as seven of 12 people, approximately 58.3%, or seven of 16 mentions, 43.75%, with the denominator stated.

Follow-up: Reconcile the per-person coding ledger before comparing percentages; preserve people with no issue and multiple themes.


## Collection protocol

When and whom to ask: Pilot the category descriptions after a specific task, before using them to code a broader response set.

Decision owner: The analyst owns the person-versus-mention denominator; the questionnaire reviewer checks category wording.

Follow-up: Recalculate both views from the same coded records and verify that totals reconcile without assuming categories are exclusive.

Response handling: The pilot uses free-text questions. Category counts are a separate manual coding example; no native multi-select, scoring or import feature is promised.


## Worked measurement study

All example records are synthetic; manual worksheet, no automatic import.

### Inspect the synthetic example

Theme | People mentioning | Share of 12 people | Share of 16 mentions
--- | --- | --- | ---
Connection | 5 | 5/12 ≈41.7% | 5/16 =31.25%
Access | 4 | 4/12 ≈33.3% | 4/16 =25%
Instructions | 7 | 7/12 ≈58.3% | 7/16 =43.75%

### Derivation

Mentions = 5+4+7=16. Respondent shares sum to 16/12×100≈133.3% because people can mention multiple themes. Mention shares sum to 100%.

Chart note: instructions appear in seven of 12 people’s answers (58.3%) and seven of 16 coded mentions (43.75%). The views answer different questions.

Calculator inputs and convention: {"mode": "ratio", "labels": ["People mentioning instructions", "People answering"], "values": [7, 12], "unit": "% people mentioning the theme", "integer": true, "subset": true}

### Evidence boundary

Observed / Three coded counts with an explicit overlap rule / Supports respondent or mention shares when labelled.
Unknown / Independent incident counts or population prevalence / Do not treat overlapping mentions as separate people.
Next check / Reconcile counts to the per-person coding ledger / Retain none and multi-theme cases.

### Method references

Pew Research Center question design: https://www.pewresearch.org/writing-survey-questions/ — Wording and response-option consistency.