# Consulting clinic: four practice studies These invented studies practise asking researchers for information before recommending an analysis. They differ from the assessed project cases. No case model or fitted answer is supplied. All conversations happen as classroom role play; do not contact real researchers. ## Roles and timing Choose a consultant, researcher and checker. At 13:05 the consultant reads the initial message and asks questions; the researcher reads the matching briefing and only supplies information requested. Keep unknown facts unknown. After eight minutes, rotate roles and use another study for eight minutes. Spend ten minutes writing a response of at most 120 words, then share an unresolved issue. ## A: milk production — initial message > We recorded daily milk production while cows received two feeding supplements in different periods. There are hundreds of rows. Which supplement works better? ### Researcher briefing Researchers first group twenty cows by early or late lactation stage. Within each stage group, they randomly assign each cow to receive A then B or B then A. This assigned order is called the sequence. Each cow has seven daily milk records in each of two measurement periods, in kg/cow/day. The schedule includes an adaptation interval. Production may change between periods even if supplement has no effect. The records do not document whether earlier supplement effects had time to disappear, whether such effects might persist, or why some daily records are missing. The researcher wants to compare mean production during the two supplements over the observed cows; the size of a useful difference still needs a reason. Ask about order, periods and earlier treatment effects before recommending an analysis. No fitting is required. ## B: lamb growth — initial message > We recorded individual lamb weight gains on two supplements. Can we treat the lambs as replicates and report the better supplement? ### Researcher briefing Researchers group twelve paddocks by three soil conditions, with four paddocks in each group. Within each soil group, they randomly assign two paddocks to each supplement. Six lambs share each paddock and receive its supplement together. Researchers record six-week gain and each lamb's weight before assigning supplements. They have not selected the main comparison or decided how baseline weight should enter the analysis. Some gain measurements are missing, and pasture quality beyond the soil groups is not documented. Researchers assign supplement to paddocks, not separately to lambs. Lambs sharing a paddock may grow alike. The outcome is gain in weight, not feed conversion. Do not invent a model formula or useful difference to complete the response. ## C: mastitis — initial message > We counted mastitis during follow-up on farms using two management approaches. Some cows had repeat episodes. Could you analyse the counts as disease risk? ### Researcher briefing Researchers select eighteen separate farms. Each farmer had chosen the farm's management approach; researchers assign no treatment. One row records a farm's clinical episode count and total cow-months observed. A cow can have more than one episode. Farms differ in cow numbers and follow-up duration. A cow-month is one cow observed for one month; it provides observation time for an episode rate. The researcher wants to compare episodes per amount of observation time, which differs from the fraction of cows ever affected. Management, herd size and season may vary together. Recording completeness, episode definitions and other farm differences need clarification. An observed management relationship does not itself show what changing management would cause. ## D: poultry cameras — initial message > Cameras gave us hundreds of activity measurements from broiler houses at different light levels. Can we show that brighter lighting increases activity? ### Researcher briefing Researchers select eight poultry houses and record six hourly mean activity summaries per house on one shared day. They record the existing light level; they do not assign brighter or dimmer lighting. Light may also change with hour. Each hourly outcome is a positive activity index calculated from many camera frames. Houses differ in flock age and management. Missing camera coverage, hourly changes and differences between houses need clarification. The question concerns how recorded light and activity vary together. More frames describe the same houses; they do not add houses or establish that changing lighting would cause the recorded activity difference. ## Write the recommendation In at most 120 words, state what comparison the researcher can investigate and in which units. Explain how researchers assigned treatments or selected groups, and which measurements share an animal or group. Describe a possible analysis in words, one check, one assumption and one essential information request. Say what information or alternative assumption might change your advice. Different justified recommendations are acceptable; no numerical answer is required.