Descriptive and Experimental Study Designs
Week 8 does not teach study design. It links to an article that does, and this page sets out that article's taxonomy, its ranking of designs by strength of evidence, and its treatment of confounding, with every claim tied to where it came from.
What you are reading, and whose it is
The design material on this page comes from a published article set as required reading in the week on quantitative research. It appeared in a sport and exercise science web journal, volume 4, issue 1, in 2000, and states its own length as 4,318 words. It is the most technically substantial document in the supplied material, and it was written for researchers who run controlled trials and crossovers on human performance; it describes its own subjects as "tissues, cells, animals, or humans".
That matters for how you use it. The subject set this reading; it did not teach it. Neither the 23 slides nor the 2,898 words of study notes reproduces the taxonomy below, and no slide reconciles the article's classification with the deck's own. Take a design distinction off this page into a project management study and you are borrowing from another discipline — a borrowing you have to justify, because the teaching material never attempts it. The subject's own account of the field is at What Quantitative Research Is.
The two classes, and what the article files under each
THE SET READING'S TABLE 1, REPRODUCED
| Class, as the table names it | Designs listed under it |
|---|---|
| "Descriptive or observational" | "case"; "case series"; "cross-sectional"; "cohort or prospective or longitudinal"; "case-control or retrospective" |
| "Experimental or longitudinal or repeated-measures", "without a control group" | "time series"; "crossover" |
| "Experimental or longitudinal or repeated-measures", "with a control group" | The table names no sub-types on this branch |
Two classes, five descriptive sub-types, three experimental arrangements. The most complete design taxonomy in anything supplied to this library, and entirely the set reading's: the teaching material contains no equivalent table.
The five descriptive designs, as the article defines them
DESCRIPTIVE DESIGNS, DEFINITIONS AS GIVEN
| Design | Definition, in the article's words | Also called |
|---|---|---|
| Case | "reports data on only one subject" | — |
| Case series | "Descriptive studies of a few cases" | — |
| Cross-sectional | "variables of interest in a sample of subjects are assayed once and the relationships between them are determined" | — |
| Cohort | "some variables are assayed at the start of a study ... then after a period of time the outcomes are determined" | prospective; longitudinal |
| Case-control | compares subjects who have an attribute against subjects who do not, on exposure to something suspected of causing it | retrospective |
| Historical cohort | a study where the difference in exposure is known but the exposure data are gathered retrospectively at a single time point | Text only, not in Table 1 |
"Descriptive studies are also called observational, because you observe the subjects without otherwise intervening." Definitions quoted or closely paraphrased; the last row appears in the article's prose, not its table.
The examples attached to these definitions are worth reading and not worth carrying: an outstanding athlete; dietary habits followed by incidence of heart disease; volume of high intensity training; alcoholic drinks per day; elite against sub-elite athletes. One example in the passage is not from that discipline — "a dysfunctional institution", offered in passing as an alternative instance of a case study and never developed. Note too that "a few cases" is the only quantity given for a case series anywhere in the source, so the boundary between a case series and a small cross-sectional study is yours to define and defend.
The experimental arrangements, and the problem each one solves
The article treats the experimental class as a sequence of fixes: each arrangement exists because the one before it leaves a specific alternative explanation open. Read in that order they are much easier to hold than as a list.
The three arrangements, in the order the article builds them
Time series, the simplest experiment
"one or more measurements are taken on all subjects before and after a treatment". A special case is the single-subject design, in which measurements are taken repeatedly on one or a few subjects before and after an intervention. The article's illustration of "repeatedly" is "e.g., 10 times", which is an example in its own discipline and not a rule.
The problem a time series leaves open
"any change you see could be due to something other than the treatment." The article's instances: subjects do better on a second test because of their experience of the first, or they change their diet between tests because the weather changed.
Crossover, one solution
Subjects are given two treatments, one real and one a control or reference. Half receive the real treatment first and half the control first; "after a period of time sufficient to allow any treatment effect to wash out, the treatments are crossed over". Any effect of retesting, or of anything that happened between the tests, "can then be subtracted out by an appropriate analysis".
Control group, for when the effect will not wash out
"If the treatment effect is unlikely to wash out between measurements, a control group has to be used." All subjects are measured; only the experimental group receives the treatment; all are measured again; and the two changes are compared.
What comes next in the article — random assignment, blinding and the placebo — is a separate set of controls on a separate set of biases, handled at Controlling Bias: Randomisation and Blinding.
The evidence hierarchy, and the asymmetry inside it
The article ranks designs on one criterion only: "the quality of evidence they provide for a cause-and-effect relationship between variables". That is worth stating out loud, because causal evidence is not the only thing a design is good for, and the ranking says nothing about any other virtue.
THE SET READING'S EVIDENCE HIERARCHY, WEAKEST TO STRONGEST
| Rank | Design | What the article says it gives you |
|---|---|---|
| 1 | Case; case series | "the weakest" |
| 2 | Cross-sectional; case-control | "good evidence for the absence of a relationship"; only "suggestive evidence of a causal connection" where a relationship is found; "a good starting point to decide whether it is worth proceeding to better designs" |
| 3 | Prospective, that is cohort | "more difficult and time-consuming to perform, but they produce more convincing conclusions about cause and effect" |
| 4 | Experimental | "the best evidence about how something affects something else" |
| 5 | Double-blind randomised controlled trial | "the best experiments" |
The ranking is the article's, in its own discipline, and it ranks on evidence for causation only.
The asymmetry at rank 2 is the most useful thing on this page and it is easy to miss. A weak design that finds nothing is telling you something reasonably solid; the same design finding something is telling you very little. That is why the article calls a cross-sectional or case-control study a good starting point rather than a poor one: it is cheap, it can rule a relationship out, and ruling out is often the decision you actually need. Nothing in the teaching material makes this point.
Confounding, and the two remedies the article offers
Confounding is the article's name for the standing hazard in any descriptive design being read causally, and it defines the term precisely: confounding occurs "when part or all of a significant association between two variables arises through both being causally associated with a third variable".
Its worked instance, from its own field and carrying no numbers, is a population study showing a negative association between habitual activity and degenerative disease. Older people are less active; older people are more diseased; so the association turns up whether or not either one causes the other.
Remedy 1: restrict the sample
- "you make sure all your subjects are the same age"
- The confounder cannot vary, so it cannot confound
- Buys precision in the estimate of the effect
- Costs applicability: the effect "generalizes only to subjects with the same narrow range of characteristics"
Remedy 2: measure it and model it
- "you include age in the analysis to try to remove its effect on the relationship between the other two variables"
- The confounder is allowed to vary and is accounted for
- Keeps the finding applicable to a wider group
- Costs precision when the sample is small; the article states a soft limit on how many such variables a small sample carries
That choice is what the article elsewhere calls the trade-off between precision and applicability, set out with its costs at Choosing What to Measure. Either way it must be settled before you collect data: remedy 1 changes who you recruit, remedy 2 changes what you record about them.
Where the deck and its own set reading disagree
The week's slide on the types of quantitative research gives a three-way split; the article it sets as required reading gives a two-way one. These are not two views of the same map, and on one design they take opposite positions.
TWO TAXONOMIES IN ONE WEEK
| The deck's types slide | The set reading |
|---|---|
| Three types: "Experimental", "Quasi-experimental (almost)", "Non-experimental (single case)" | Two classes: "Descriptive or observational" and "Experimental or longitudinal or repeated-measures" |
| "Non-experimental (single case) — continuous assessment of behaviour over a period of time that involves intervention effects over time" | The same design, repeated measurement on one or a few subjects before and after an intervention, is the single-subject design, a special case of the time series, filed under Experimental |
The deck offers no sub-types, no ranking and no crosswalk to the article, and its term "quasi-experimental" has no counterpart there. The article offers five descriptive sub-types, three experimental arrangements and a five-level evidence hierarchy, and no crosswalk to the deck. Neither document acknowledges the other.
Reading your own design off the taxonomy
This is the part you can use tomorrow, provided you keep it inside its limits. The classification is a classification, not a standard: applying it tells you what your design can support, not that your design is adequate. The judgements below apply the article's logic to project settings; the article does not make them, and neither does the teaching material.
What your design lets you claim
What to carry forward
- The taxonomy, the hierarchy and the definition of confounding here belong to one article published in a sport and exercise science journal in 2000. They are not the subject's positions and they are not standards.
- Descriptive designs establish association; experiments establish causation. The article's body qualifies both halves, and the qualifications are the useful part.
- A weak design that finds nothing is more informative than the same design finding something. That asymmetry appears nowhere in the teaching material.
- Confounding has two remedies here: restrict the sample, or measure the confounder and include it in the analysis. Both are decided before data collection, not after.
- The deck and its own set reading file the single-case design on opposite sides of the experimental line. Report the disagreement; do not pick a winner for the reader.
Frequently asked questions
Can I cite this design taxonomy in a project management dissertation?
You can cite it as a published article, provided you attribute it and note that it was written for sport and exercise science. What you cannot do is present it as the subject's teaching or as a general standard. The supplied teaching material offers a different, three-way taxonomy and never reconciles the two.
Is a case study the same as a case series?
Not in this taxonomy. A case "reports data on only one subject"; a case series covers "a few cases". No number or upper bound is given for "a few", so the boundary between a case series and a small cross-sectional study is undefined in the source.
Why does the article call a cross-sectional study a good starting point when it also calls it weak?
Because the weakness is one-sided. Such a study "can provide good evidence for the absence of a relationship", so it can rule things out cheaply. Only when it finds a relationship does the evidence drop to suggestive, and that is the point at which the article says it is worth proceeding to a stronger design.
What does the article mean by longitudinal?
Two different things, and it says so. Under descriptive designs it is a synonym for a cohort or prospective study; it is also the article's synonym for the whole experimental class. The article flags the ambiguity itself, so any proposal using the word should state which sense is meant.
Does the teaching material teach any of this?
No. The week sets the article as required reading and its own slides give a different classification, with no crosswalk in either direction. Study design is not named in any of the week's three stated learning objectives.
How do I handle confounding if I cannot restrict my sample?
The article's second remedy applies: measure the suspected confounder and include it in the analysis. That decision has to be made before you collect data, because it determines what you record about each unit. There is a cost in precision when the sample is small.
References and source attribution
- Hopkins 2000, 'Quantitative Research Design', a sport and exercise science web journal, vol. 4, no. 1. The article states its own length as 4,318 words and carries a notice pointing to an updated version published in 2008; the version set as the week's required reading, and the version cited throughout this page, is the 2000 one.
- The supplied teaching source: the week 8 slide deck on the nature of quantitative research, in particular the slide defining the three types of quantitative research and the activity slide that sets the article above as required reading.
- The supplied teaching source: the week 8 study notes, in five numbered sections, two of which reproduce part of the article above verbatim. The design taxonomy on this page is not among the reproduced material.
Suggested questions for Ask KEVOS
- Classify my study design using the set reading's taxonomy and tell me what claim it supports.
- Draft the limitations paragraph for a cross-sectional project survey that found a positive association.
- List the confounders I should consider for a study relating team size to schedule variance, and say for each whether to restrict or to model it.
- Explain the difference between a crossover and a control group design in terms I can put to a sponsor.
- Where does the teaching material contradict the set reading on study design, and how should I write around it?