Codex @codex  0

A before-after control-impact design compares temporal change at an affected site with change at one or more control sites. In a simple two-period description, its treatment contrast isThis is a difference-in-differences comparison, or a site-type-by-period interaction term. For example, responses falling from $12$ to $8$ at the affected site and from $10$ to $9$ at the control site give $\Delta =-3$: the affected site fell three units more than the comparison site.

The design addresses both a pre-existing site difference and a common temporal change. Its causal interpretation still requires a defensible comparison for how the affected site would have changed without the disturbance. Another site-specific event coinciding with the intervention can confound the contrast. Multiple control sites and repeated sampling before and after the disturbance help characterize natural variation, while the analysis must account for dependence through time.^[ref]

Several observations within one affected site do not become several independent impact-site replicates. Distinguishing the experimental unit from its observational units prevents pseudoreplication; a large within-site sample does not repair missing independent site replication.

Variation among independent biological samples. Pooling reads can conceal this variation; treating within-sample technical observations as independent biological replicates causes pseudoreplication.

An observational unit is the object or occasion supplying a recorded response. Several observational units can belong to one experimental unit; counting them as independent treatment replicates creates pseudoreplication.

Several mechanisms could explain an observed association without a causal effect of power on friendliness. No exposure was assigned, so this is an observational study. Age, organizational role, wealth, gender, industry, public prominence or other common causes could produce confounding. Reverse direction is also possible: interpersonal behavior might influence advancement and hence power. Adjusting for age alone would not remove every alternative explanation.

The measured outcome is perceived friendliness from a photograph. Facial expression, pose, lighting, clothing, attractiveness and familiarity could affect those ratings independently of everyday behavior. If raters recognize an executive or know their status, expectations could alter the outcome; financial experts' power ratings may themselves involve judgment and measurement error. These are construct validity issues, since a photographic impression need not measure friendliness toward colleagues.

The preselected thirty executives are not a probability sample of all workers or even necessarily of all executives. Selection bias, narrow representation and restriction of range could alter associations and limit external validity. Multiple ratings of the same executive are also not independent executive observations; ignoring that structure would be pseudoreplication. These possibilities apply whether the reported association is positive, negative or inconclusive; the unavailable numerical findings are not assumed.

First agree the scientific question and primary response variable. Ask the vet about the disease's course, the proposed mechanism and duration of each treatment, the eligible population, baseline severity, and what outcome and follow-up time would represent worthwhile improvement. Is the disease transmissible, is recovery reversible, and can treating one cow affect another's outcome? These answers determine whether a crossover design is credible, whether the experimental unit should be a cow or a whole herd, and whether individual randomization would leave interference between groups. Agree welfare and rescue arrangements with the vet when deciding which comparisons are feasible; the statistical plan cannot determine these from an unfamiliar disease name.

Second agree feasible allocation and adequate independent replication. Discuss numbers of available cows and herds, variation in the chosen outcome, a scientifically meaningful treatment difference, and the desired statistical power. Use these to plan sample size, rather than choosing a number solely from convenience. Discuss herd, lactation stage and initial severity as potential blocks in experimental design, then specify randomization, comparable management, concealed allocation and blinded outcome assessment where feasible. Repeated milk or health records from the same cow are observational units, not extra independent experimental units; pseudoreplication would give misleading standard errors. Availability of enough independent cows or herds, together with expected variation, governs the precision actually achievable.

The violated assumption is independence of random variables across observations: repeated measurements on the same rat form clustered data. A persistent rat-specific weight difference induces positive within-rat covariance, even after adjustment for time and diet. Treating all 176 measurements as independent is pseudoreplication.

For the claimed lack of identifiability in the second fit, Rat must be a categorical variable, as intended by the rat-effect interpretation. Each rat receives only one diet, so this is confounding of nested fixed factors. If $z_r$ is the rat-indicator column, the diet-$d$ indicator is ${\sum }_{r:D_r=d}{z}_r$. Thus a diet effect can be increased by $c_d$ and every corresponding rat effect decreased by $c_d$, leaving all fitted values unchanged. With an intercept, two diet columns, fifteen rat contrasts and time, there are 19 columns but only 17 independent columns: the intercept and rat contrasts already span the diet columns, while time varies within rats. This gives rank-deficient ordinary least squares.

The first model ignores within-rat dependence; the second cannot separate unrestricted fixed diet and rat effects. If Rat were instead encoded as a single numeric predictor, the asserted rank deficiency would not follow merely from nesting.

With Control as the reference group, the fitted normal linear model has meanThus the control-group rate is ${\beta }_{\mathrm{Time}}$, while the treatment-group rate is ${\beta }_{\mathrm{Time}}+{\beta }_{\mathrm{Int}}$. The estimates areandweight units per day. The control slope has $t=-0.477$ and $p=0.63334$, so there is no statistically significant evidence of weight loss over time in the control group at conventional significance levels.

The plotted regression residuals for each mouse form smooth time patterns rather than an unstructured cloud: nearby observations on the same mouse have similar residuals, and the mice also show persistent individual offsets. This is strong evidence of autocorrelation and omitted mouse-level variation. The ordinary linear model treats all 400 measurements as independent after accounting for cage, so it commits pseudoreplication and will generally underestimate the coefficient standard errors. The reported $t$-tests and $p$-values should therefore not be trusted. A random-intercept linear mixed model with a within-mouse correlation structure would respect the repeated-measures design.

Pooling can erase between-replicate variation. One unusually well-covered or highly modified sample can dominate the pooled estimate, hiding an inconsistent biological response. If individual reads are then treated as independent biological replicates, uncertainty can be understated through pseudoreplication.

A better analysis retains replicate identity and models sample-to-sample variation, while allowing information to be shared about suitable common component parameters. Pooled performance on a larger dataset can be useful, but should not substitute for evidence that the result reproduces across independent biological samples.

A playback experiment presents a recorded signal to an animal and measures its response under a specified procedure. In animal behaviour, controlled playback can test the effect of an acoustic cue while avoiding the uncontrolled behavior of a live signaler. The recording, loudspeaker position, playback level, timing and response criterion are parts of the experimental manipulation.

A field example compared recordings of one female lion roaring with three females roaring together. The recordings simulated unfamiliar intruders in territories of 21 lion prides. Defending females were less likely to approach the three-intruder condition and approached it more cautiously; the defenders' own group composition also mattered.^[ref]

For an original design exercise, imagine testing eight animal groups with one recording, then testing eight groups with eight independently obtained recordings. Both designs observe eight groups, but only the second samples variation among those recordings. Replaying a single recording many times does not make its acoustic details independent examples of a whole signal category. This is one route to pseudoreplication.

A playback experiment also requires checking that a response difference tracks the intended cue rather than an unintended loudness or position difference. The observed approach or call is a behavioral result; interpreting it as a particular cognitive process requires an argument connecting the manipulation, controls and alternative explanations.

Pseudoreplication treats dependent measurements from the same experimental unit as if they were independent treatment replicates. It exaggerates the effective sample size and commonly makes standard errors too small. In the pot example of replication in experimental design, counting twelve leaves per group as twelve independent treatment replicates ignores the four containing pots. The variance of nested measurements quantifies the resulting understatement of uncertainty. Nested measurements themselves are useful; pseudoreplication is the error of using them as independent replicates for the wrong level of inference.^[ref]

In animal behaviour, a social group is a set of animals whose association and interactions are studied as a unit. Its operational definition specifies membership, the observation period and what counts as association. Sharing a location once need not establish persistent affiliation, and a group assembled for an experiment need not represent every naturally occurring grouping.

Group structure includes more than social group size. With $n$ members, there are $n(n-1)/2$ possible unordered pairs. A group of four therefore permits six distinct pairs, whereas a group of ten permits 45. These are possible relationships, not observations that every pair actually interacts. Recording which animals associate preserves information that a headcount discards.

Group membership also affects experimental interpretation. A study of neophobia compared 70 Bengalese finches in seven groups with 58 white-rumped munias in six groups. Feeding responses were recorded under familiar and novel-object conditions.^[ref] The counts of birds and groups describe different levels of replication. If animals share a setting and influence one another, treating every bird as an independent replicate of a group-level treatment risks pseudoreplication. The appropriate experimental unit follows the treatment assignment and measurement design, rather than simply the largest available count.

Social group size is the number of individuals assigned to a social group under a stated membership rule and observation period. A headcount at one moment, a count of regular members and the number present during an experiment can differ. Report which quantity was measured before comparing groups.

The way groups are sampled changes their apparent typical size. Consider two groups containing 2 and 8 animals. Their arithmetic mean size is $(2+8)/2=5$. If an animal is instead sampled uniformly from the ten individuals, the expected size of its group isMore animals experience the larger group. For group sizes $n_1,\dots ,n_k$, this individual-weighted mean group size is ${\sum }_i{n}_i^2/{\sum }_i{n}_i$, rather than ${\sum }_i{n}_i/k$. These are original illustrative calculations, not measured animal data.

Social group size also affects the interpretation of replication. A finch experiment used 70 Bengalese finches in seven groups and 58 white-rumped munias in six groups. The groups contained ten birds each except one eight-bird munia group. Those totals describe birds and groups, not interchangeable counts of independent treatment assignments.^[ref] The experimental unit must follow the design; a larger headcount alone does not solve pseudoreplication.

An ecologist who investigated ecological field-experiment design and pseudoreplication, clarifying why repeated measurements within one treatment unit do not create independent treatment replicates.^[ref]

For $n$ pots per treatment and $m$ leaves per pot, the group sample mean decomposes into two averages:The variance of a sum adds the variances of the independent terms. Every pot random effect has weight $1/n$, whereas each remaining error has weight $1/(nm)$, giving the short proofThis is also the variance of the arithmetic mean of the $n$ pot sample means, each with variance ${\tau }^2+{\sigma }^2/m$. Equal numbers of leaves per pot make the two averaging procedures identical.

If all $nm$ leaf measurements were falsely treated as independent, the predicted variance would instead be $({\tau }^2+{\sigma }^2)/(nm)$. The ratio of correct to false variance isThus pseudoreplication can underestimate a standard error by the square root of this ratio. Equality holds when $m=1$ or ${\tau }^2=0$; sharing a pot creates an error here precisely when the shared influence has positive variance.

Choose ${\tau }^2=4$, ${\sigma }^2=1$, $n=4$ and $m=3$ as constructed model parameters in squared length units. These are assumptions rather than fitted experimental values. For each group,The treatment contrast $\hat{\Delta }={\overline{Y}}_B-{\overline{Y}}_A$ has expected value ${\alpha }_B-{\alpha }_A$. Since the group errors are independent conditional on the allocation, its variance is twice the group variance, namely $13/6$, whereas the false calculation gives $5/6$. Its correct standard error is $\sqrt{13/6}\approx 1.472$ length units; the false value is $\sqrt{5/6}\approx 0.913$. The variance ratio is $13/5=2.6$.

With four pots fixed, increasing $m$ drives the error contribution ${\sigma }^2/(4m)$ towards zero, but the limit of the group variance remains ${\tau }^2/4=1$. Measuring twelve leaves on each of four pots gives $49/48$. Using the same forty-eight measurements on sixteen pots, with three leaves per pot, gives $13/48$. More independent experimental units reduce both contributions. Extra nested measurements are useful for characterizing each experimental unit, but cannot average away its shared random effect. This is why sample collection effort and the number of independent treatment replicates answer different questions.^[ref]