Learn · Guides

Reading a Wright map

The Wright map (after Ben Wright) is the single most useful picture a Rasch analysis produces: your respondents and your items, plotted on the same logit ruler. One glance answers questions that tables bury — is the instrument aimed at the people taking it? Where does it measure precisely, and where is it blind?

The layout

One vertical scale, two sides. The person distribution (usually the left side) shows where your sample sits — each person’s measure comes from their total score, mapped through the model. The item side shows each question at its calibrated difficulty. High on the map means more of the trait (more able, harder); low means less. In Logit the map is interactive: zoom into a crowded region, hover an item for its exact measure, and export the view for a paper.

The three questions it answers

1. Targeting — do the distributions line up? Measurement is most precise where items are dense near the people. If the person distribution floats a logit above the items, your instrument is too easy for the sample: everyone succeeds on nearly everything, and you learn little that distinguishes them. (A mismatch isn’t always a flaw — a screening test should concentrate items near its cut score, not the population mean. Judge targeting against purpose.)

2. Gaps — where is the ruler missing markings? A vertical stretch with people but no items is a region where the instrument can’t discriminate: everyone in the gap gets nearly the same score. Gaps are the to-do list for the next round of item writing — they tell you exactly what difficulty to write for.

3. Redundancy — are items stacked? Five items at the same difficulty measure that point well, but four of them may be costing respondent time without adding range. Wide, even item spread is the ideal ruler.

A worked example

In the Liking for Science walkthrough, the map shows “go on a picnic” anchoring the bottom (almost every child likes it — it barely measures) while “find bottles and cans” tops the scale (only the most science-attached children like it). The vertical span between them is the construct, made visible: what liking science means, from easy affection to demanding enthusiasm. Checking that this ordering matches theory is the first validity argument an instrument makes.

Numbers that go with the picture

Person separation / reliability summarize what the map shows: separation is roughly “how many distinct levels of the trait can this instrument tell apart in this sample” (Logit reports the strata index (4G+1)/3 in its PDF). Low separation usually looks, on the map, like a tight person clump against a sparse item side. Item standard errors shrink with sample size and with how close the item sits to the people — another reason targeting matters.