Denoised Correlation Matrix

v1.0.0

Compute a denoised asset correlation matrix, using one of the following methods:

The eigenvalues clipping method, described in the first reference, which is based on random matrix theory

References

Input

{
  "type": "object",
  "required": [
    "assets",
    "assetsCorrelationMatrix",
    "assetsCorrelationMatrixAspectRatio"
  ],
  "properties": {
    "assets": {
      "type": "integer"
    },
    "assetsCorrelationMatrix": {
      "type": "array",
      "description": "assetsCorrelationMatrix[i][j] is the correlation between the asset i and the asset j",
      "items": {
        "type": "array",
        "items": {
          "type": "number"
        }
      }
    },
    "assetsCorrelationMatrixAspectRatio": {
      "type": "number",
      "description": "The aspect ratio of the asset correlation matrix, defined as the number of assets divided by the number of asset returns per asset used to compute the asset correlation matrix"
    },
    "denoisingMethod": {
      "type": "string",
      "description": "The method used to denoise the asset correlation matrix",
      "enum": [
        "eigenvaluesClipping"
      ]
    }
  }
}

Output

{
  "type": "object",
  "required": [
    "assetsCorrelationMatrix"
  ],
  "properties": {
    "assetsCorrelationMatrix": {
      "type": "array",
      "description": "assetsCorrelationMatrix[i][j] is the correlation between the asset i and the asset j; assetsCorrelationMatrix is possibly null in case the denoising method did not manage to denoise the provided asset correlation matrix",
      "items": {
        "type": "array",
        "items": {
          "type": "number"
        }
      }
    }
  }
}