The analysis of ‘moderation’, ‘interaction’, ‘mediation’ and ‘longitudinal growth’ is widespread in the human sciences, yet subject to confusion. To clarify these concepts, it is essential to state causal estimands, which requires the specification of counterfactual contrasts for a target population on an appropriate scale. Once causal estimands are defined, we must consider their identification. I employ causal directed acyclic graphs and single world intervention graphs to elucidate identification workflows. I show that when multiple treatments exist, common methods for statistical inference, such as multi-level regressions and statistical structural equation models, cannot typically recover the causal quantities we seek. By properly framing and addressing causal questions of interaction, mediation, and time-varying treatments, we can expose the limitations of popular methods and guide researchers to a clearer understanding of the causal questions that animate our interests.

The human sciences should seek generalisations wherever possible. For ethical and scientific reasons, it is desirable to sample more broadly than ‘Western, educated, industrialised, rich, and democratic’ (WEIRD) societies. However, restricting the target population is sometimes necessary; for example, young children should not be recruited for studies on elderly care. Under which conditions is unrestricted sampling desirable or undesirable? Here, we use causal diagrams to clarify the structural features of measurement error bias and target population restriction bias (or ‘selection restriction’), focusing on threats to valid causal inference that arise in comparative cultural research. We define any study exhibiting such biases, or confounding biases, as weird (wrongly estimated inferences owing to inappropriate restriction and distortion). We explain why statistical tests such as configural, metric and scalar invariance cannot address the structural biases of weird studies. Overall, we examine how the workflows for causal inference provide the necessary preflight checklists for ambitious, effective and safe comparative cultural research.

Causal inference requires contrasting counterfactual states under specified interventions. Obtaining these contrasts from data depends on explicit assumptions and careful, multi-step workflows. Causal diagrams are crucial for clarifying the identifiability of counterfactual contrasts from data. Here, I explain how to use causal directed acyclic graphs (DAGs) to determine if and how causal effects can be identified from non-experimental observational data, offering practical reporting tips and suggestions to avoid common pitfalls.

Confounding bias arises when a treatment and outcome share a common cause. In randomised controlled experiments (trials), treatment assignment is random, ostensibly eliminating confounding bias. Here, we use causal directed acyclic graphs to unveil eight structural sources of bias that nevertheless persist in these trials. This analysis highlights the crucial role of causal inference methods in the design and analysis of experiments, ensuring the validity of conclusions drawn from experimental data.

A polled sheep with a short tail, and devoid of wool on the head, legs, belly and breech, has been proposed as a breeding goal (Scobie et al 1997). This paper examines the practical and welfare implications of such an objective. Mean shearing times were significantly (P < 0.001) slower for control Romney (97s) and Coopworth (88s) ewe hoggets compared to times for both breeds after they were ‘trimmed’ to resemble the breeding goal (47s). Romney wethers were slower to shear than ewes (P < 0.001), but trimmed wethers (53s) were much faster to shear than the untrimmed controls (99s). Trimmed sheep produced less wool (pooled mean weights of fleece wool for trimmed sheep vs total wool for controls being 2.5 vs 3.5 kg; P < 0.001), half the difference (ie 500g) consisting of lower value oddment wools from the untrimmed controls. Tails were docked to 0, 20, 40, 60, 80 and 100 per cent of the distance between the base of the tail and the hock of Perendole lambs, and 0, 20, 40 and 60 per cent on Coopworths. More dags accumulated as the tail stump increased in length, although significance was dependent on the time of year and management system. No flystrike (cutaneous myiasis) was observed in Perendales given physical and chemical protection, while a small proportion of all tail lengths were flystruck in Coopworths given only physical protection. When shorn as hoggets, the time taken to shear Perendole ewes increased with increasing tail length (P < 0.001). Provided the undocked short tail is bare of wool, the proposed breeding goal should deliver an economic solution to sheep breeders attempting to satisfy markets conscious of chemical residue and animal welfare issues in the new millennium.

Although Vector Autoregressive models are commonly used to forecast prices, specification of these models remains an issue. Questions that arise include choice of variables and lag length. This article examines the use of Forecast Error Variance Decompositions to guide the econometrician's model specification. Forecasting performance of Variance Autoregressive models, generated from Forecast Error Variance Decompositions, is analyzed within wholesale chicken markets. Results show that the Forecast Error Variance Decomposition approach has the potential to provide superior model selections to traditional Granger Causality tests.