Descriptive and Diagnostic Designs
Chapter -Eight
Syllabus topic 7, "Preparation of the Research Design"
Pages 462 to 464 of 543
In one line
A descriptive design establishes what is the case and a diagnostic design establishes what goes with what, and unlike the exploratory design both must be fixed completely before a single unit is studied.
In the wording a student can write in an exam: the descriptive research design is used to portray accurately the characteristics of a defined population or situation, and the diagnostic design to determine the frequency with which something occurs or the extent to which it is associated with something else; both are rigid rather than flexible, requiring the population, the sample, the instrument and the plan of analysis to be settled in advance, since their purpose is to produce comparable data on which magnitude and association can be established.
The descriptive design
Its question. What is the case, in what proportion, how often, among whom.
Its requirements, and every one is a decision taken in advance.
The population defined exactly, chapter 1000.
The sample drawn by a stated method, and by a probability method if the results are to be generalised, chapter 1010.
The instrument fixed and pre-tested, chapter 970, and unchanged thereafter, because a change mid-collection produces two incomparable datasets.
The operational definitions settled, chapter 1050.
The analysis planned, including the tables, chapter 1200.
And the treatment of non-response decided, chapter 990.
Its rigidity is the point. An exploratory design changes as it learns; a descriptive design must not, because its whole value lies in every unit having been treated the same way.
The diagnostic research design
Its question. What is associated with what, and how often does something occur in relation to something else.
How it differs from the descriptive design. In practice very little in method: the same population, sampling and instrument requirements apply. The difference is in the question and therefore in what the design must anticipate.
What it must anticipate additionally. The variables it will relate must both be recorded, and there must be enough units in each combination to permit comparison, chapter 1000. A design that records whether the accused was informed but not the size of the sitting cannot answer a diagnostic question about the two, chapter 1200's empty tables test.
And the standing caution. Association is not causation, chapter 1050, so a diagnostic design establishes that two things go together and not that one produces the other.
In much writing the two are treated as one category, and a student may say so: descriptive-diagnostic designs share their requirements and differ in the question asked.
What both require that the exploratory design does not
| Exploratory | Descriptive and diagnostic | |
|---|---|---|
| Purpose | Formulate the problem | Establish magnitude or association |
| Flexible | Yes, deliberately | No, fixed before collection |
| Sampling | Purposive | Probability, where generalisation is intended |
| Instrument | Unstructured, may change | Fixed and pre-tested |
| Product | A hypothesis | A finding |
| May report a proportion | No | Yes |
Descriptive and Diagnostic Designs
The elements of a descriptive design, as a checklist
A student can write these eight lines and have a design.
One. The population, stated exactly, with the place and the period.
Two. The sampling frame, and what it omits.
Three. The sampling method and the size, with the reason for each.
Four. The instrument, attached as an appendix.
Five. The operational definitions.
Six. Who will collect, and how they will be trained if it is not the researcher alone.
Seven. The plan for editing, coding and tabulation, with the empty tables.
Eight. The expected limitations.
And the test: hand those eight lines to another student and ask whether they could run the study. If they could not, the design is incomplete, chapter 1190.
The commonest failures
Changing the instrument mid-collection, which produces two studies.
Substituting respondents, which destroys the sample, chapter 1000.
Recording only what the hypothesis needs, so that no unexpected finding is possible. A little redundancy in the instrument is cheap insurance.
Not deciding the analysis in advance, which permits selection after the fact, chapter 1190.
And treating an exploratory study as a descriptive one when writing up, chapter 1210, which is where the percentages that should never have been computed come from.
A worked example
The two studies this book has used, classified.
The awareness study, chapter 1000, is descriptive. Its question is what proportion of adult respondents in three settlements have heard that a lawyer can be provided free of cost. Population, frame, systematic sampling, fixed schedule, operational definition in the words of question 5, planned tabulation, and refusals recorded.
The remand study, chapter 1200, is descriptive as to H1 and diagnostic as to H2. H1 asks the proportion informed, which is description. H2 asks whether the proportion varies with list length, which is diagnosis, and it requires the schedule to record both variables and the analysis to have enough sittings in each size band.
And the clinic register study, chapter 990, is descriptive and is a census, so it needs no sampling design at all and can make a complete statement about that population.
Three studies, all within a student's reach, and each one's design follows from its question, which is chapter 1190's whole argument.
Quick revision
Descriptive design: portrays accurately the characteristics of a defined population or situation. Diagnostic design: determines frequency of occurrence or the extent of association with something else.
Both are rigid. Population, sample, instrument, operational definitions, analysis and treatment of non-response are all settled before collection, because their value lies in every unit being treated the same way.
Descriptive and Diagnostic Designs
The diagnostic design additionally requires that both variables be recorded and that there be enough units in each combination.
Eight-line checklist: population with place and period; frame and what it omits; sampling method and size with reasons; instrument as an appendix; operational definitions; who collects and their training; the plan for editing, coding and tabulation with the empty tables; and the expected limitations.
The test: could another student run the study from your design?
Failures: changing the instrument mid-collection; substituting respondents; recording only what the hypothesis needs; not planning the analysis; and writing up an exploratory study as though it were descriptive.
Test yourself
1. Distinguish the descriptive from the diagnostic design. The descriptive design portrays the characteristics of a defined population or situation, answering what is the case and in what proportion. The diagnostic design determines the frequency of occurrence of something or the extent to which it is associated with something else. Their method requirements are largely the same; the difference lies in the question and in what the design must therefore anticipate.
2. Why must both be rigid, unlike the exploratory design? Because their value depends on every unit having been treated in the same way, so that the data are comparable and a magnitude or an association can be established. An instrument changed part way through produces two incomparable datasets rather than one.
3. Give the eight lines of a descriptive design. The population with place and period; the sampling frame and what it omits; the sampling method and size with reasons; the instrument, attached as an appendix; the operational definitions; who will collect and how they are trained; the plan for editing, coding and tabulation with the empty tables; and the expected limitations.
4. What must a diagnostic design anticipate that a descriptive one need not? That both of the variables to be related are recorded by the instrument, and that there will be enough units in each combination of them to permit a comparison; otherwise the association cannot be examined however good the rest of the design.
The rest of this subject
These notes are cut from the University's printed syllabus. Open the syllabus itself, or the past papers, for the same subject.