MICA fills the missing cells using a Bayesian model that respects positive definiteness, anchors each missing cell to what the surrounding matrix implies, and tells you cell by cell which imputed values you can trust.
A partial matrix is a correlation matrix where some cells have observed values and other cells are blank. In the MASEM literature this is also called an "incomplete" matrix. (Note: it's not the same as a "partial correlation," which is a different statistical concept.)
Easiest: a CSV with a header row and a row-name column. Diagonal = 1 (or blank). Missing cells blank or "NA".
Also accepted: Excel (.xlsx) with the same shape; whitespace-separated numbers; lower-triangle text (one row per line, length growing 1, 2, 3, …); a square text block of numbers with no names. If we can't parse it deterministically and you've enabled the LLM fallback on the server, we'll try that.
Symmetry: if you only fill the upper triangle (or only lower), we'll mirror it for you. The diagonal is forced to 1.