‘Table of rows’ in SQL Databases equivalent to ‘Collection of documents’ in MongoDB. And Star Data Model can be represented as document with nested properties, like this:

{
  _id,
  totalA,
  totalB,
  dimentionA: {
    _id,
    dataA1,
    dataA2,
    ....
  },
  dimentionB: {
    _id,
    dataB1,
    dataB2,
    ....
  }
}

To put simply, your fact document will contain dimension documents as nested objects.
With this structure, you will not need to use any joins ($lookups), because every dimension is already joined to the fact data.

This scheme will be a perfect fit, only if the total size of such document will not exceed 16MB.

Check this OracleDB and MongoDB Comparison for more detailed overview.