Q methodology

tickStat is a survey and experiment platform that runs complete Q methodology studies online. It is built by academic researchers, for academic researchers, and Q methodology is one of the disciplines it was designed to support natively. The full Q-sort workflow lives on a single platform: designing the Q-set and the consensus pyramid, the drag-and-drop sorting interaction, and a one-click Factor Analysis with varimax rotation and publication-ready interpretation reports.

Q methodology is a workhorse method in policy research, environmental psychology and stakeholder viewpoint analysis — wherever the question "what shared positions exist in this group?" matters more than "how many people hold each position?". Respondents Q-sort a set of statements onto a forced-distribution pyramid, and factor analysis identifies the latent viewpoints that organise the sorted data.

Your first Q study in five steps¶

Sign in with Google. You get a free Silver licence with every methodology included (up to 10 surveys and 40 completed respondents per survey).
Add the pre-sort. Create a matrix question with your statements (the Q-set) as rows and three columns such as disagree, neutral and agree.
Add the Q-sort. Create a Q methodology question, link it to the matrix and define the pyramid: how many columns it has and how many statements fit in each one.
Preview and field the study in up to 14 languages. The pyramid is designed for desktop and tablet screens.
Run the analysis. Open Analysis → Q-Model Analysis: the consensus pyramid first, then the Factor Analysis with the factor arrays, the distinguishing and consensus statements and the interpreted report.

Q methodology in brief¶

Q methodology studies subjectivity: the viewpoints that exist in a group about a topic. Instead of asking many people a few questions and counting answers, it asks a deliberately chosen group to rank a whole set of statements against each other, and then looks for people who ranked them in a similar way. A study has six parts, and each one maps to a step in tickStat:

Concourse. Everything that is said about the topic: interviews, literature, media, policy documents. It is collected before the study and is the source of the statements.
Q-set. The sample of statements drawn from the concourse, usually a few dozen, chosen to cover the range of opinions rather than to represent them proportionally. In tickStat they are the rows of the pre-sort matrix and the cards of the pyramid.
P-set. The participants. They are selected because they are likely to hold different and relevant views, not as a random sample, so Q studies work with small groups.
Q-sort. Each participant ranks the statements under a condition of instruction (for example, sort the statements from those you most disagree with to those you most agree with) onto a forced, roughly bell-shaped distribution. In tickStat this is the drag-and-drop pyramid, preceded by the pre-sort into three groups.
Factor analysis. The Q-sorts are correlated person by person and factor-analysed. Each factor gathers participants who sorted the statements alike: a shared viewpoint. tickStat extracts the factors by principal components with varimax rotation and flags the participants who define each one.
Interpretation. Each factor is described from its factor array (the sort of an ideal participant who represents it), from the statements that distinguish it from the other factors and from the statements all factors agree on. tickStat shows the three and drafts an interpreted report.

Why tickStat for Q methodology¶

Drag-and-drop sorting on the pyramid itself. Participants pre-sort the statements into three groups and then drag each card into the forced-distribution pyramid, moving cards between cells until the sort reflects their view. The whole Q-sort runs in the browser, on desktop and tablet, inside the same survey as the rest of your questions.
Consensus Pyramid dashboard. Visualise the forced-distribution structure (column counts, statement positions) in real time as the data comes in. Spot pyramid violations, monitor study balance and decide when you have enough data to estimate factors.
One-click Factor Analysis with varimax rotation. Eigenvalues, factor loadings, defining sorts and factor scores computed inside the platform — no round-trip to a separate stats package required for the first cut. Export the underlying per-respondent sorts for any custom analysis you want to run downstream.
Factor interpretation report, ready for publication. The platform generates a textual summary of each factor (consensus statements, distinguishing statements, defining sorts) that drops straight into the results section of a paper.
Reproducible and open. Q-set, instructions, sorting interface, factor extraction settings — every parameter is exposed and documented. Reproducing a study or sharing it with a co-author means sharing the survey definition, not a screenshot of a wizard.
Multilingual fielding in 14 languages. Run the same Q study across countries with a single survey definition.
GDPR and EU AI Act compliant. Important for academic researchers running publication-grade studies under the current regulatory regime.

What's on this page¶

The rest of this page documents the workflow in tickStat: how the Q-sort works for the respondent, how to configure the pre-sort matrix and the forced-distribution pyramid, the Consensus Pyramid and Card Distribution dashboard, the Factor Analysis (how the factors are extracted and how to read the factor arrays and the distinguishing and consensus statements), how to segment the analysis by respondent groups, and the reports and exports.

If you are new to the platform, start with the Getting started guide.

This question type implements Q methodology: each respondent sorts a set of statements (the Q-set) onto a pyramid-shaped grid, and the analysis groups respondents who sorted the statements in a similar way into shared viewpoints (factors). A Q study in tickStat uses two questions: a matrix where the respondent first classifies the statements into broad groups, and the Q question itself, where the statements are placed on the pyramid.

How the Q-sort works for the respondent¶

The sort is done in two stages, the usual procedure in Q methodology:

Pre-sort. The respondent reads every statement in a matrix and classifies it into a few broad groups, typically disagree, neutral and agree.
Pyramid. The statements appear as cards, already grouped into baskets according to the pre-sort. The respondent moves each card from its basket into a cell of the pyramid, from the columns at one end (for example most disagree) to the columns at the other end (most agree). The number of cells in each column is fixed, so the respondent has to prioritise: only a few statements fit in the extreme columns.

In the pyramid the respondent can drag a card from any position to any other position, including cards already placed, so the sort can be revised until it reflects their view. Clicking a yellow card sends it to the bottom of its stack, which lets the respondent look through the cards of a basket without placing them.

Screen size. The pyramid needs room: it shows every column and every card at once. Plan the study for desktop or tablet screens and tell participants so in the invitation. On a phone the grid is too large to sort comfortably, so it is not recommended.

Configuration¶

Step 1 — the pre-sort matrix. Create a matrix with radio-type columns. Each row is a statement of the Q-set and each column is one of the broad groups (for example disagree, neutral, agree). The answer titles in the matrix should match the basket labels of the Q question.

Step 2 — the Q question. Create the Q methodology question and configure:

Matrix Segmentation Question: The source matrix configured previously.
Rows: Number of rows in the pyramid.
Columns: Number of columns in the pyramid.
Direction of the matrix: Accepted values are "up" or "down," depending on the desired display orientation.
Literal baskets: Titles for the source bins where items are initially presented for the user to move into the pyramid. There is one input box per basket — the number of boxes follows the number of bins defined in Source of the bags (three by default). Example: En desacuerdo, Neutras, De acuerdo.
Literal columns: Text displayed in the pink rectangles, with one input box per column (the number of boxes follows the Columns field). Leave a box empty to show no header for that column.
Detailed column literal: Text displayed below the pink rectangles, also one input box per column. Leave a box empty to show no detail for that column.
Source of the bags: This important configuration links matrix columns to Q model bins. For example, 1,2;3;4,5 means that matrix items selected as answer 1 or 2 go to the first bin, items selected as answer 3 go to the second bin, and items selected as answer 4 or 5 go to the third bin.

The label boxes are generated automatically from the configured counts: change Columns (or the number of bins in Source of the bags) and the corresponding set of boxes is regenerated, keeping the values you already typed. To omit a label, simply leave its box empty.

Step 3 — the shape of the pyramid. After saving the configuration, a pyramid based on the specified rows and columns appears, where you need to set the items for each column. This defines the forced distribution of the study: how many statements the respondent can place in each column. The total number of cells must equal the number of statements in the Q-set, so that every statement has exactly one place on the pyramid.

Visual Results Dashboard¶

Q Model questions include an integrated visual results dashboard, accessible from the Analysis menu (the Q-Model Analysis entry, which appears whenever the survey contains a Q-Model question). If the survey has a single Q-Model question the dashboard opens directly. If it has several, you first see a selection screen where you tick the Q-Model question or questions to analyse (use Select all to tick them all) and press Open analysis:

One question — the dashboard analyses that question on its own, exactly as before.
Several questions (combined) — the selected questions are pooled into a single analysis: their Q-sorts are merged into one matrix and you get a single consensus pyramid, card distribution and factor analysis over the combined data, with the header showing the combination (e.g. "P52 + P55 (combined)"). This is meant for surveys where the same instrument is repeated (for example the same statements sorted into an "up" and a "down" pyramid). The selected questions must share the same set of statements; if they don't, the selection is rejected with a message and you are returned to the selection screen. The pyramid orientation (up/down) is purely visual and does not affect the analysis, so up and down variants pool cleanly.

The dashboard offers three analysis modes: Consensus Pyramid, Card Distribution and Factor Analysis. The first two describe the sample as a whole and are available as soon as responses come in; Factor Analysis identifies the distinct viewpoints.

Consensus Pyramid¶

Shows an aggregate view of where all respondents placed each card on average. Cards are color-coded by consensus level: - Green (High consensus, σ < 1.3): Most respondents placed the card in the same position. - Yellow (Medium, σ 1.3–1.5): Some variation in placement. - Red (Low/Polarized, σ > 1.5): Respondents disagree significantly about this card.

Hover over any card to see its exact mean position, standard deviation, and number of respondents.

The consensus pyramid is a description of the whole sample, not a viewpoint: a statement in red is one that respondents placed far apart, which is usually a statement that will separate the factors in the Factor Analysis.

Card Distribution¶

Click any card from the left panel to see a horizontal bar chart showing how respondents distributed that card across all columns. The list is sorted by consensus level, making it easy to identify the most agreed-upon and most polarized items.

Which respondents are included¶

The dashboard counter, the Factor Analysis and the Interpreted Report all consider only respondents who completed the entire survey (status Done). Pyramid placements from respondents who finished the Q Model question but later abandoned the survey are excluded so that statistics and factor groupings are not distorted by partial data.

Factor Analysis¶

The Factor Analysis tab runs a Q-methodology factor analysis using R (requires the qmethod R package). In Q methodology the analysis is by person: respondents, not statements, are correlated with each other, and each factor is a group of respondents who sorted the statements in a similar way — a shared viewpoint.

Running the analysis¶

Select the number of factors: a fixed number from 2 to 7 (3 by default), or Auto (Kaiser criterion).
Click Run Analysis. The system computes the correlations between respondents, extracts factors via PCA with varimax rotation, and identifies distinguishing and consensus statements.
Results show factor profiles (Q-score pyramids per factor), the number of respondents per factor, and variance explained.
The Statements Analysis tabs show consensus statements (items all groups agree on) and distinguishing statements per factor (items that differentiate each group).
Click Generate Interpreted Report (Word) to produce a Word document (.docx) with an AI-generated interpretation of the results, including factor profile images and data tables, or Download user scores (Excel) to get the factor loadings of every respondent.

There is no limit on the number of respondents of a Q study. The Factor Analysis uses every respondent with a complete Q-sort (all the cards placed on the pyramid). Only in very large studies, with more than 3,000 complete Q-sorts, it runs on a random sample of 3,000 to keep the computation fast; in that case a notice above the results states how many were used ("Analysis based on a random sample of N respondents out of M total").

How the factors are extracted¶

Correlation and extraction. The Q-sorts of the respondents are correlated with each other and the factors are extracted by principal components analysis (PCA).
Rotation. The factors are rotated with varimax, which makes each respondent load as clearly as possible on a single factor.
Number of factors. With Auto, the system keeps the factors whose eigenvalue is greater than 1 (Kaiser criterion), with a minimum of two. With many respondents this criterion tends to keep more factors than can be interpreted, so treat Auto as a starting point: run the analysis with one factor fewer and one more, and keep the solution whose factors are all interpretable and are each defined by several respondents.
Defining respondents (flagging). A respondent defines a factor when their Q-sort is clearly associated with that factor and not with the others. The qmethod package flags them automatically: the loading on the factor must be statistically significant (p < 0.05) and its square must be larger than the sum of the squared loadings of that respondent on all the other factors. Respondents who load on several factors, or on none, are not flagged and do not contribute to the factor profiles.
Factor scores. For each factor, a z-score per statement is computed as the weighted average of the positions given by its defining respondents (respondents with a higher loading weigh more). The z-scores are then converted to the column values of the pyramid: the result is the Q-score of each statement on that factor.

Reading the results¶

Summary cards. The first card shows the number of factors. Then there is one card per factor with the number of respondents that define it and the percentage of variance it explains. A factor defined by very few respondents is a weak viewpoint: consider a solution with fewer factors.

Factor Profiles (Q-scores per factor). One column per factor, listing every statement with its Q-score, from the highest (the statements that viewpoint agrees with most, in the positive columns of the pyramid) to the lowest. This list is the factor array: the sort of an ideal respondent who represents the viewpoint perfectly. It is the main material for interpreting and naming each factor.

Statements Analysis. A table with the Q-score of each statement on every factor, split into tabs:

Consensus: statements whose score does not differ significantly between any pair of factors. They are the common ground of all the viewpoints.
Distinguishing F1, F2, …: statements whose score on that factor differs significantly from its score on the other factors. They are what makes that viewpoint different, and the place to start when writing its description.

The Type column shows the classification of each statement (for example, whether it distinguishes one factor only or all of them).

A common reading order is: check the summary cards to see whether every factor has enough defining respondents, read the factor profile of each factor from both ends towards the centre, then use the distinguishing statements to describe what separates each viewpoint and the consensus statements to describe what they share.

Segmenting the analysis¶

Respondent filter bar: Above the dashboard there is a collapsible Respondent filters bar that lets you segment the analysis by demographic or screening answers (e.g. age range, region, gender, study group). Pick one or more filterable questions, choose the values you want to keep, and press Apply. The Consensus Pyramid, the Card Distribution and the Factor Analysis all recompute on the matching subset, and the stats card updates to show how many respondents are included. Clear all restores the full Done dataset. Whenever you change the filter the previous Factor Analysis results are cleared so you re-run the analysis on the new subset, and the Interpreted Report (Word) generated afterwards describes the same filtered subset. Available filter questions: radio, checkbox, single-choice and predefined-list (listaValores) demographic items. In addition, three respondent-metadata dimensions — Source, OS and Browser, the same ones shown as Respondent segments in the Interactive Dashboard — appear at the end of the question picker, so you can segment the Q analysis by entry panel, operating system or browser family. They are available once the survey's dashboard data has been synced (press Refresh Data on the Interactive Dashboard if they don't show up yet); their values are the actual buckets observed in the data.

Reports and exports¶

Interpreted Report (Word). Generated from the Factor Analysis results with the Generate Interpreted Report (Word) button. It contains an AI-generated interpretation of each factor together with the consensus pyramid, the card statistics, the factor profile images and the data tables. Review it as a first draft: the interpretation of a Q study is the researcher's responsibility.
User scores (Excel). The Download user scores (Excel) button downloads one row per respondent with the respondent identifiers and their loading on each factor (the correlation between that respondent's Q-sort and the factor). Use it to see which respondents belong to each viewpoint and to cross the factors with other answers of the survey. It is available after a successful Run Analysis.
Raw Q-sorts (Complete Report). The Excel Complete Report includes the Info qModel tab, with one row per card placed by each respondent: survey number, user identifier, question identifier, column, row, card identifier, statement text and category. This is the data to use if you want to run the analysis in another tool.

Definitions (Info qModel tab): - Columna (Column): Matrix column index. - Fila (Row): Matrix row index. Note: Rows are numbered from bottom to top.