Skip to main content

Statistical Consulting: Information to Obtain at the First Meeting

Statistical Consulting: Information to Obtain at the First Meeting

Your consultant will be using the following document as a guide at the first meeting. It is important that you read this document in order to be prepared for the types of questions you will be asked. Do not worry about understanding the details, as it is written for the consultant and may contain statistical terminology that is unfamiliar to you. Part I is for a designed experiment and part II is for a survey study.

I. A typical designed experiment

Modify as needed for the particular study

1. Background and objectives

  • Motivation and previous results. Pilot study?
  • Specific research hypotheses.

Ask the client: What would an answer look like? A drawing, diagram or graph may help.

  • Scope of Inference Narrow (Fixed) vs Broad (Random).
    e.g. Interest in just these blocks or beyond?

Useful material: Appropriate section(s) of research proposal.

2. Treatment structure

  • Nature of treatments: Factorial structure? Controls? Reasons for treatment levels? Fixed and/or random factors?
  • What purpose do the controls serve? How is the client interested in using the controls?
  • Any observational factors? Populations of interest
  • Crossed vs. Nested

3. Design structure

  • What are the Experimental Units (e.u.'s)? Method of selection. Intended vs. actual scope of inference? More than one size of e.u. (e.g., split unit design)? For observational factors: What are the sampling units?
  • Subunits or groupings? (Animals in pens, fish in tanks, subplots, subsampling, etc.)
  • Any Blocking? Gradients? Patterns in variation? Natural block size? Time as blocks?
  • Blocks: Fixed or Random?
  • Assignment of e.u.'s to treatments. (Randomized? Systematic? Restrictions on randomization?) For observational factors: How were sampling units selected?
  • Replication. Amount? At what level(s) of the design?
  • Standard design? (Ex: Split plot with completely randomized design at whole plot level)

At design stage: Resources available, time budget, tradeoffs between more measurements per e.u. versus more e.u.'s.

Useful material: Diagram/map of experimental layout, "Methods" section of proposal.

4. Response structure

  • How are responses obtained? Steps needed.
  • Nature of responses. (Nominal, ordinal, interval, ratio, etc.)
  • Subsampling? Identify measurement units as distinguished from experimental units.
  • Repeated measures on the same e.u.? (In time, space or along some other gradient?)
  • Auxiliary measurements? Purpose? (Increase precision? Explanatory variables? Other?)
  • Data issues. (e.g., missing data? Outliers? Skewness? Heavy‑tails? etc.)

Useful material: Graphs, e.g., scatter plots, or tables showing actual or hypothetical data

5. Assistance needed

  • Specific question(s) that client would like consultant to address.

Timing: When is answer needed? (e.g., presentation date, thesis defense date, etc.)

  • Statistical background of client. (e.g., courses, statistical software, etc.)

Useful material: Example(s) of computer output


II. A typical survey study

Modify as needed for the particular study

1. Background and objectives

  • Motivation and previous results.
  • Specific research hypothesis. What would an answer look like? Would a drawing, diagram or graph help to explain? Is interest in simple population description or formal comparison between some groups?

Useful material: Appropriate section(s) of research proposal.

2. Sampling Design

  • Consulted with Survey Research Center (SRC) on overall survey design and/or implementation of a survey?
  • Population(s) of interest. Target vs. sampled population. Intended vs. actual scope of inference.
  • What are the sampling units? Sampling frame? Census?
  • Pilot Study? If already done, what was learned? Information for defining strata?
  • Strata or sub‑populations? Are estimates of interest at this level? Comparisons between?
  • Clusters? (Sampling unit is group of smaller units, e.g., household as cluster of individuals)
  • Sample Selection. How selected? Probability sampling? (e.g., simple random, systematic, etc.)
  • General type of sampling design. (SRS, stratified random sampling, multi‑stage, etc.)
  • Level of Non‑Response? Plan for non‑response? (Follow‑ups, etc.)

At design stage: Resources available, time budget, tradeoffs between more measurements per population unit versus more population units in sample.

Useful material: Maps, subset of sampling frame.

3. Response/Measurement Design

  • Method of measurement. Telephone interview? Self‑administered questionnaire? etc.
  • Questionnaire development. New or established instrument? Assessment of validity/reliability? Pretest? Consulted with SRC on questionnaire construction?
  • Incomplete questionnaires? Expected level? Plan for?
  • Nature of responses. (Nominal, ordinal, Likert scale, interval, ratio, etc.)
  • Auxiliary measurements? Purpose? (Increase precision? Explanatory variables? other?)
  • Repeated measures on the same unit? (In time, space or along some other gradient?)
  • Data issues. (Missing data? Outliers? Skewness? Heavy‑tails? etc.)

Useful material: Copy of questionnaire, graphs/tables showing actual or hypothetical data.

4. Treatments/Interventions

E.g., pre‑post study of effects of viewing an educational video

5. Assistance needed

  • Specific question(s) that client would like consultant to address.
  • Timing. When is answer needed? (e.g., presentation date, thesis defense date, etc.)
  • Statistical background of client. (e.g., courses, statistical software, etc.)

Useful material: Example(s) of computer output.


III. Combination of I and II

A typical biological or other resource survey has some features of both I and II. Usually no questionnaire and not directed toward humans. Typically, these surveys are similar to designed experiments but with terminology similar to surveys. So instead of experimental units there are sampling units. There also tend to be factors, but not necessarily treatments. (Different habitat types could be considered factors, yet these are not treatments.)

Examples: plant ecology studies, stream studies, studies of bridges, etc.