Revision notes

Assessment and Evaluation: Practice

The key points, rules to remember, tables and diagrams from all 9 lessons, on one page for quick revision. To keep a copy, use your browser's Print and choose "Save as PDF".

Lesson 1

Classifying assessment scenarios: purpose beats timing

Four roles, by purpose
RoleQuestion it answersTypical momentExample
PlacementWhere should this learner start?Before teachingA readiness test to group new students
FormativeHow is learning going, and what should change?During teachingExit slips used to plan the next lesson
DiagnosticWhy does this learner keep making this error?When a difficulty persistsA step-by-step test on fractions
SummativeWhat has been achieved overall?At the endAn end-of-term test for the report card

Ask what the results are used for: where to start is placement, adjusting teaching is formative, the cause of a stubborn error is diagnostic, recording achievement is summative. Learners steering themselves is assessment as learning.

  • Classify by purpose, never by timing alone.
  • Diagnostic goes deeper than formative and targets persistent errors.
  • Self-assessment is 'as learning' only when learners use it to steer their learning.
  • Placement, formative and diagnostic uses are assessment for learning; summative use is assessment of learning.
Read the full lesson

Lesson 2

Types of reliability and validity in scenarios

Reliability by procedure
TypeProcedureWhat it shows
Test-retestSame test given twice to the same group, after a gap; scores comparedStability over time
Parallel (equivalent) formsTwo comparable versions given to the same group; scores comparedEquivalence of forms
Split-halfOne test given once, divided into two halves (for example odd and even items); halves comparedInternal consistency
Inter-rater (inter-scorer)Two or more markers score the same answers; their scores comparedAgreement between markers
Validity by question
TypeQuestionTypical evidence
Content validityDo the items fairly cover the syllabus and objectives?Comparing items with a blueprint
Construct validityDoes the test measure the intended trait, such as creativity or reasoning?Evidence that scores behave as the trait should
Criterion-related: concurrentDo scores agree with another measure taken now?Comparing with teachers' current ratings
Criterion-related: predictiveDo scores forecast later performance?Comparing entrance scores with later results
Face validityDoes the test look suitable to those taking it?Opinions of test-takers; weakest evidence

Same test twice: test-retest. Two versions: parallel forms. One sitting split in two: split-half. Two markers: inter-rater. Coverage of syllabus: content validity. Forecasting later results: predictive validity.

  • Name reliability types from the procedure used.
  • Content validity is the key concern for classroom achievement tests.
  • Predictive validity forecasts later performance; concurrent validity agrees with a present measure.
  • Face validity is appearance only and is the weakest evidence.
  • Very low reliability rules out high validity.
Read the full lesson

Lesson 3

Bloom's taxonomy: classifying questions at speed

Cognitive levels
Order (lowest to highest)Original (1956)Revised (2001)
1KnowledgeRemember
2ComprehensionUnderstand
3ApplicationApply
4AnalysisAnalyse
5SynthesisEvaluate
6EvaluationCreate

Revised order: Remember, Understand, Apply, Analyse, Evaluate, Create. Nouns became verbs; Synthesis became Create and moved to the top; a knowledge dimension was added.

  • Original: Knowledge, Comprehension, Application, Analysis, Synthesis, Evaluation.
  • Revised: Remember, Understand, Apply, Analyse, Evaluate, Create.
  • In the revision Create is the highest level; in the original Evaluation was.
  • Domains: cognitive, affective, psychomotor.
  • Classify by verb, then check what thinking the task truly needs.
Read the full lesson

Lesson 4

Writing and choosing items: MCQs, essays, rubrics and portfolios

Two rubric types
RubricHow it scoresBest for
AnalyticScores each criterion separately, then combinesDetailed feedback; diagnosis of strengths and weaknesses
HolisticGives one overall score from a description of the wholeQuick overall judgement of many pieces

Write MCQs with a clear stem, one best answer, plausible distractors and no clues. Match the item to the thinking required. Analytic rubrics give scores per criterion; holistic give one score. Process portfolios show growth; showcase portfolios show best work.

  • Clues such as grammar or option length lower validity.
  • Distractors should come from real errors.
  • Essays suit organisation and argument; performance tasks suit skills.
  • Analytic rubric: criterion by criterion; holistic rubric: one overall judgement.
  • Process portfolio: growth over time; showcase portfolio: best work.
Read the full lesson

Lesson 5

Norms, criteria, grades and percentiles: avoiding traps

Two ways to grade
Grading typeHow grades are setReference
AbsoluteFixed mark ranges, decided in advance (for example 80 and above = A)Criterion
RelativeGrades depend on position in the group (for example top tenth = A)Norm

Compared with others: norm-referenced (ranks, percentiles, relative grading). Compared with a standard: criterion-referenced (mastery, absolute grading). Percentage is marks earned; percentile rank is the share of the group scoring below.

  • Norm-referenced suits selection; criterion-referenced suits mastery and teaching.
  • Percentile rank = number scoring below ÷ number in group × 100.
  • Absolute grading uses fixed ranges; relative grading depends on the group.
  • Grading reduces focus on tiny mark differences but loses detail.
  • CCE as an idea means varied, continuous, whole-child assessment, not more tests.
Read the full lesson

Lesson 6

Measurement, assessment and evaluation: terms and scales of measurement

ScaleWhat the numbers doExampleFair summary measure
NominalName or label categories; no orderRoll-number labels, school houses, genderMode
OrdinalGive order, but gaps are not equalRanks, grades A to E, satisfaction levelsMedian
IntervalEqual gaps, but zero is only a conventionTemperature in degrees Celsius, test scores treated as equal stepsMean
RatioEqual gaps and a true zeroHeight, weight, time taken, distanceMean and ratios

Measurement gives a number; assessment gathers and interprets evidence; evaluation judges worth and decides. Scales rise in power: nominal, ordinal, interval, ratio.

— Rule to remember
  • Measurement is the number; evaluation is the value judgement and decision.
  • Assessment is wider than measurement and may be qualitative.
  • Nominal names, ordinal orders, interval has equal steps, ratio has a true zero.
  • Ranks and grades are ordinal, so their gaps are not equal.
  • Test marks are usually treated as interval data, since a score of zero does not mean zero knowledge.
Read the full lesson

Lesson 7

Describing test scores: mean, median, mode, range and standard deviation

MeasureHow to find itBest used when
MeanAdd all scores and divide by the number of scoresScores are fairly symmetrical with no extreme values
MedianArrange in order; take the middle score, or the average of the two middle scoresThere are extreme scores, or the data are ordinal
ModeThe most frequent score; there may be none or more than oneYou want the most common value or the data are nominal

Mean is pulled by extremes; median is the middle and resists them; mode is the most common. Positive skew: mean above median. Negative skew: mean below median.

— Rule to remember
  • Choose the median when a few scores are extreme.
  • Range uses only two scores; SD uses all of them.
  • SD is zero only when every score is identical.
  • Positive skew means a right tail and a mean above the median.
  • Equal means do not mean equal classes: always check spread.
Read the full lesson

Lesson 8

Item analysis: difficulty index, discrimination index and distractors

IndexFormulaWhat it tells you
Difficulty index (facility value, P)(U + L) ÷ 2nProportion answering correctly; high means easy, low means hard
Discrimination index (D)(U − L) ÷ nHow well the item separates upper from lower scorers
Choices on one item
OptionUpper groupLower group
A15
B00
C (key)116
D34

Difficulty = (U + L) ÷ 2n. Discrimination = (U − L) ÷ n. A negative D means the item is faulty and a distractor nobody picks is useless.

— Rule to remember
  • A higher P means an easier item.
  • D shows how well an item separates strong from weak students.
  • Negative D: check the key and wording.
  • Replace distractors that nobody chooses.
  • Use the group sizes given in the question.
Read the full lesson

Lesson 9

Observation tools, feedback and marker errors

ToolWhat it recordsTypical use
ChecklistWhether a behaviour or step is present or absent (yes or no)Did the child wash hands, measure first, then pour?
Rating scaleThe degree or frequency of a behaviour on a scaleNever, sometimes, often, always
Anecdotal recordA short, factual description of a significant incidentOne event in the playground, with date and place
SociometryWho chooses whom as a friend or partnerSeeing the pattern of group preferences
Cumulative recordInformation over the years in one fileTracking growth across classes
ErrorWhat happens
Halo effectA general impression (neat writing, good behaviour) colours marks on specific features
Leniency errorMarks are generally too generous
Severity errorMarks are generally too harsh
Central tendencyThe marker avoids high and low marks and gives everyone the middle
Contrast effectA script is judged by comparison with the one just before it

Checklist: present or absent. Rating scale: how much or how often. Anecdotal record: facts, not judgements. Halo: impression colours the marks.

— Rule to remember
  • Choose the tool to fit the behaviour: presence, degree or incident.
  • Keep facts and interpretation apart in anecdotal records.
  • Good feedback is specific, timely and points forward.
  • Name the marker error from the pattern of marks.
  • Rubrics, anonymity and moderation raise reliability.
Read the full lesson

Privacy · Terms · Refunds · Delivery