Item Response Theory (IRT) is primarily designed for categorical or ordinal data, like responses to test items scored 0/1 (incorrect/correct) or Likert-type scales. It doesn’t directly handle continuous data in its standard form.

If your dataset contains continuous values like (0, 1.751, 2.310, 0.875…), applying standard IRT models directly won’t work, because IRT assumes discrete item responses. You would need to either:

So yes, in principle, you can estimate parameters with an IRT approach, but only after either transforming the data or using a continuous-response IRT model.

Yes, IRT can deal with binary, poltymous and continuous data.

See my earlier book for an example of how to do it.

Roche, A., Wainer, H. & Thissen, D. (1975). Skeletal maturity: The knee joint as a biological indicator. New York: Plenum.

Howard Wainer

Yes, you *can*, but let me clarify a few key points before you try applying **Item Response Theory (IRT)** to your dataset.

What IRT needs

IRT is designed for item-level response data, typically in the form of:

Binary data

Polytomous data.

From such data, IRT estimates item parameters (difficulty, discrimination, guessing) and person parameters .

2. Your dataset shape

You showed a single row:

`(0, 1.751, 2.310, 0.875, …)`

This looks like continuous values, not discrete item responses. If these values represent:

Scores/responses from one individual across items → You’ll need many individuals’ responses to estimate IRT parameters.

Already-scaled values (like factor loadings, abilities, or scores) → Then they are not raw item responses, and IRT parameter estimation is not directly possible from just this row.

3. Requirements for IRT estimation

To estimate IRT parameters, you need:

A matrix rows = individuals, columns = items

4. Answering your question

If your dataset is just one row like `(0, 1.751, …)`, you cannot estimate IRT item parameters.

If your dataset is many rows (persons) × many items, and responses are categorical (binary/polytomous), then yes, you can apply IRT to find item parameters.

Here's a note I wrote about continuous Rasch models in 2001 - https://www.rasch.org/rmt/rmt144a.htm - but my experiments indicated that data is rarely truly continuous. Data that is claimed to be continuous is usually granular or lumpy. It is more meaningful to analyze the grains or lumps as ordinal categories.