Assessment Domain - Best Practices

Owned by Eric Jansson
Aug 09, 2022
5 min read

Contents

The Ed-Fi assessment specification is purposefully non-restrictive when it comes to the types of result data than can be sent to the Assessment Outcomes API, so that vendors with different types of data can be accommodated. While differences exist between assessment platforms, there are enough similarities to warrant a suggested method of returning data to the API for the most common use cases. These suggested methods will help assure consistent data types for the most common use cases. This document addresses six of those use cases, for typical, point-based assessments.

An assessment’s overall maximum and minimum score

For a typical point-based assessment’s overall maximum and minimum score we suggest an AssessmentReportingType descriptor value of "Number score" and a ResultDataType descriptor value of "Decimal".

The score elements should reflect the total number of points available. We recommend you round to two decimal places which are always present. For instance, for an assessment that has a maximum of 50 points, the suggested formulation for the “scores” element in json is:

"scores": [
    {
        "assessmentReportingMethodType": "Number score",
        "resultDatatypeType": "Decimal",
        "maximumScore": "50.00",
        "minimumScore": "0"
    }
]

An assessment’s performance levels

For a typical point-based assessment’s performance levels we suggest an AssessmentReportingMethodType descriptor value of "Raw score" and a ResultDataType descriptor value of "Percentage". We recommend no decimal places.

A unique consideration with performance levels is that the level cutoffs can be either inclusive or exclusive. We suggest that the MinimumScore be inclusive while the MaximumScore not be considered part of the performance band. In the example json shown below, this would indicate that a score of 80 would be considered “Proficient”. In addition, if it is possible for your students to achieve better than 100% on a performance level, we suggest that a MaximumScore value for the top performance level be omitted.


"performanceLevels": [
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Advanced",
               "minimumScore”: "90",
               "resultDatatypeType”: "Percentage”
           },
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Proficient",
               "minimumScore”: "80",
               "maximumScore": "90",
               "resultDatatypeType”: "Percentage”
           },
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Basic",
               "minimumScore”: "70",
               "maximumScore": "80",
               "resultDatatypeType”: "Percentage”
           },
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Below Basic",
               "minimumScore”: "0",
               "maximumScore": "70",
               "resultDatatypeType”: "Percentage”
           }
       ],


An objective’s scores designation

For a typical point-based assessment objective’s scores designations we suggest an AssessmentReportingType descriptor value of "Raw score" and for the ResultDataType descriptor value we recommend "Percentage", as shown in the example json below.


"scores": [
      {
          "assessmentReportingMethodType": "Raw score",
          "resultDatatypeType": "Percentage"
      },

An objective’s performance levels

For a typical point-based objective’s performance levels we suggest an AssessmentReportingMethodType descriptor value of "Raw score" and a ResultDataType descriptor value of "Percentage". We recommend no decimal places.

A unique consideration with performance levels is that the level cutoffs can be either inclusive or exclusive. We suggest that the MinimumScore be inclusive while the MaximumScore not be considered part of the performance band. In the example json shown below, this would indicate that a score of 80 would be considered “Proficient”. In addition, if it is possible for your students to achieve better than 100% on a performance level, we suggest that a MaximumScore value for the top performance level be omitted.


"performanceLevels": [
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Advanced",
               "minimumScore”: "90",
               "resultDatatypeType”: "Percentage”
           },
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Proficient",
               "minimumScore”: "80",
               "maximumScore": "90",
               "resultDatatypeType”: "Percentage”
           },
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Basic",
               "minimumScore”: "70",
               "maximumScore": "80",
               "resultDatatypeType”: "Percentage”
           },
           {
               "assessmentReportingMethodType": "Raw score",
               "performanceLevelDescriptor": "http://namespace.com/Below Basic",
               "minimumScore”: "0",
               "maximumScore": "70",
               "resultDatatypeType”: "Percentage”
           }
       ],

A student’s final score for the overall assessment

For a student’s final score on a typical point-based assessment we suggest an AssessmentReportingType descriptor value of "Number score" and for the ResultDataType descriptor value we recommend "Decimal". 

For the result, we recommend you round to two decimal places which are always present. The result should reflect the raw score of the student’s final result. For instance, for a score of 37 out of 50 points, the result would be shown as 37.00. 


"scoreResults": [
    {
        "assessmentReportingMethodType": "Number score",
        "result": "37.00",
        "resultDatatypeType": "Decimal"
    },
],

A student’s final score on an objective

For a student’s score on a typical point-based objective’s performance levels we suggest an AssessmentReportingMethodType descriptor value of "Raw score" and a ResultDataType descriptor value of "Percentage".

For the result we recommend two decimal places, reflecting the percent correct the student achieved for the objective.


"scoreResults": [
   {
       "assessmentReportingMethodType": "Raw score",
       "result": "42.86",
       "resultDatatypeType": "Percentage"
   }
 ],

Assigning Assessment Identifiers

When assigning identifiers for Assessment, ObjectiveAssessment, and Assessment metadata and StudentAssessment

“Identifiers” were introduced into the assessment domain as partial surrogate keys to allow the source systems employed by assessment organizations to use their internally unique identifiers instead of adhering to a natural key system created by the education organizations.  A namespace was created to ensure that the identifiers created by an assessment organization were unique within a district or state ODS.  

The original intent was to allow Assessment to be identified by:

  1. An internal key for the assessment, potentially even a computer-generated number, that is used to populate the data into the API.
  2. A namespace reflecting the uri for the assessment organization.

This pattern of an Identifier +Namespace was introduced for Assessment, ObjectiveAssessment, and Assessment metadata and StudentAssessment.

However, assessment results are often being obtained from files that provide the assessment results and may not directly contain metadata about the assessment or objective assessments.  These files are loaded into the ODS in batch mode.  These loaders must create these identifiers by convention and thus there is great amount of variance and in some cases misunderstanding on how best to do this.  The is a temptation to create complex patterns of identifiers – but this is not necessary.

To best understand, it is important that one consider the key structures, as follows:

EntityComposite Key
Assessment
  • AssessmentIdentifier
  • Namespace
ObjectiveAsessment
  • ·Assessment reference
    • AssessmentIdentifer
    • Namespace
  • IdentificationCode
AssessmentItem
  • ·Assessment reference
    • AssessmentIdentifer
    • Namespace
  • IdentificationCode
StudentAssessment
  • Assessment reference
    • AssessmentIdentifer
    • Namespace
  • Student reference
    • StudentUniqueID
  • StudentAssessmentIdentifier

For Assessment Identifier

The composite key for Assessment is:

  1. Namespace is the uri of the source assessment organization. Thus, for the ACT assessment, the namespace would be “uri://act.org”
  2. AssessmentIdentifer needs to unique identify the specific assessment within the context of the assessment organization, consisting of:
    • Internal ID for the assessment by the assessment organization, or
    • Short identifier for the assessment that is unique within the organization across the various assessments they offer. For example, there could be “ACT” and “ACT Practice” assessments that are offered.
    • Optionally a string representing the version of the assessment, as it may be pertinent such as:
      • Version number, such as 3.2; or
      • Year or school year, such as 2021
    • NOT the assessment organization, since this is reflected in the namespace

For ObjectiveAssessment IdentificationCode

The composite key for ObjectiveAssessment is:

  1. Namespace is the uri of the source assessment organization. Thus, for the ACT assessment, the namespace would be “uri://act.org”
  2. AssessmentIdentifer is the unique identifier for the specific assessment within the context of the assessment organization.
  3. IdentificationCode for the ObjectiveAssessment needs to uniquely reflect the “subtest” in the context of the AssessmentIdentifier and the Namespace.
    • Internal ID for the objective assessment by the assessment organization, or
    • Reflect the topic of the objective assessment, for example Mathematics, Reading, Science, Writing, etc.
    • NOT the assessment or assessment organization, since that is reflected in the AssessmentIdentifer and the Namespace.

For AssessmentItem IdentificationCode

The composite key for AssessmentItem is:

  1. Namespace is the uri of the source assessment organization. Thus, for the ACT assessment, the namespace would be “uri://act.org”
  2. AssessmentIdentifer is the unique identifier for the specific assessment within the context of the assessment organization.
  3. IdentificationCode for the AssessmentItem needs to uniquely reflect the item in the context of the AssessmentIdentifier and the Namespace. This is typically either:
    • Internal ID for the assessment item by the assessment organization, or
    • A unique identifier of the item from the test bank; or
    • A generated number, potentially even a sequence number

For StudentAssessment IdentificationCode

The composite key for StudentAssessment is:

  1. Namespace is the uri of the source assessment organization. Thus, for the ACT assessment, the namespace would be “uri://act.org”
  2. AssessmentIdentifer is the unique identifier for the specific assessment within the context of the assessment organization.
  3. StudentReference consisting of the StudentUniqueID
  4. StudentAssessmentIdentifier for the student’s results from an Assessment. The only uniqueness requirement is to distinguish between multiple times the student takes the same assessment.  
    • Possible values:
      • A booklet (if paper) or session ID (if electronic). or
      • The assessment vendor’s ID for the student, or
      • The ID for the student obtained from rostering, or
      • An internal ID for the student’s results for this administration
    • NOT the assessment or assessment organization, since that is reflected in the AssessmentIdentifer and the Namespace.
    • NOT the Ed-Fi StudentUniqueID since it is already part of the key.