Jeromy Anglim's Blog: Psychology and Statistics


Tuesday, September 22, 2009

Statistics Consulting: My Approach and Orientation

I've been a statistical consultant for postgraduate and fourth year psychology students for many years. The following post distils my method and philosophy of statistics consulting.


Structure of the session:
  • Face-to-face meeting: Sessions are structured as face-to-face meetings in my office.
  • Email is for scheduling the meeting, not discussing issues: Most of my consulting sessions start with a client sending me an email seeking to make an appointment. This is an effective use of email. However, I try to avoid discussing statistical issues via email. There are several reasons for this: 1) the presenting problem is rarely the actual problem; 2) the amount and style of information delivery can not be adapted to the client's comprehension; 3) most of the tools of effective communication (e.g., notepad, statistics software) are not present; 4) there is little scope for back and forth communication
  • Session duration: A one-hour session is usually appropriate to talk through a research project. If more time is needed, it is typically better to reschedule for at least a week later. This gives the client time to absorb what has been discussed.
  • Follow up email: If I have identified follow-up points, I send these links through by email.
Standard Elements of a Session
  • Scene setting: This involves recording key elements of the study:  Research design (sample size, sample type, design, experimental manipulation, measures, main variables, etc.); educational level of the researcher (fourth year, masters, PhD, etc.); research context (supervisors, supervisor expectations);  research questions (general; and operationalised). Clients sometimes need guidance about what information a consultant needs in order to advise. Two important things to correct are: a) the client who is immersed in their research who assumes too much domain specific knowledge on the part of the consultant; b) the client who gives domain specific detail which is not relevant to the data analytic decisions.
  • Discussion of presenting problem: The client usually has a particular concern, which led to the initiation of contact. I discuss this with the client. This is also a chance to take notes on other issues that should be discussed.
  • Outlining the results section: The aim of this process is to assist the client to develop a basic outline of their results. It involves a discussion of what analyses to present and what order to present them in. It may also involve a discussion of  what tables and figures to present.
Recording the session on Paper:
  • Overview: A fundamental tool that I use in a consulting session is the A4 Pad of Paper. It is used to record notes during the session. The pad is placed on a desk and positioned to enable both of client and consultant to see and write on the pad. The rationale for the pad is to: 1) aid the working memory of both the client and consultant; 2) record the discussion for future reference; and 3) allow for visual and other novel representations. The following types of information are recorded:
  • Top of the page: Client Full Name and Date.
  • Core information: sample size (by group, time point, etc. as necessary); design (observational or experimental; cross-sectional or longitudinal); record main variables; who they are working with.
  • Issues: As a client speaks about their research project, I identify issues. I record these as dot points in a separate section of the pad. Discussion of issues can then be deferred, so that the client's flow of thought is not disrupted.
  • Follow up material: Dot points are recorded regarding follow up material that I want to send to the client after the session. This might be a blog link or references to journal articles or book chapters.
  • Blog posts: If I identify an issue that may be relevant to other researchers, I'll flag this and write a post on it. This post is then potential follow-up material for future clients.
  • Procedures: A procedure can be set out in dot-point form. It is useful for the consultant to say to the client that the consultant will outline the procedure quickly on paper, and then explain it, giving details and guidance about where to find additional information.
  • Discussing models: a loose version of SEM notation with circles and arrows can be useful for discussing relationships between variables. Model notation (e.g., y ~ a + bx) can also be good.
  • Outlining figures and tables: When I suggest that a client should consider presenting results in terms of a table or figure, it is useful to draw an outline of the table and figure.
  • Copies: The client usually takes a copy as a record of the session. I typically ask the client to photocopy the notes and put them in my pigeon hole to provide a record of the session. This can then be used in any future sessions.
Common General Advice
General Orientation
  1. Client decides: The consultant provides resources to help the client learn and make their own decisions. The consultant clarifies questions and provides information on the pros and cons of available options.
  2. Encourage proactivity: The consultant encourages a proactive orientation in the client. The consultant highlights how the client can learn about statistics, such as reading specific books or searching the internet. This is reinforced during the session by keeping the client active in the discussion process, and regularly asking questions. It is NOT the role of a consultant to do the analyses for the client. In general, it is typically more appropriate to provide details of an accessible reference that the client can read than to spend too long in a session explaining details.
  3. Goal directed: A consulting session is goal directed.
  4. Search for underlying issues: The presenting problem is usually only part of the issue. A presenting problem usually provides the opportunity to discuss other statistical issues that the client may not have thought about.
  5. Sensitivity and standards: Clients sometimes experience conflicts when analysing and writing up their results. These include: showing technical prowess using advanced analytic procedures versus being clear and simple; presenting flaws versus hiding flaws (such as violation of important assumptions, low reliabilities, tests that seemed not to work, and so on); present exploratory findings in a confirmatory way versus in an exploratory way. Consulting balances sensitivity with an honest assessment of the situation. Consulting brainstorms ways of analysing and reporting data with integrity that also meet some or all of the client's needs.
  6. Individualised treatment: The consultant understands that clients come with diverse backgrounds. In particular clients vary in prior experience and the importance they attribute to their current analysis needs and the development of data analytic skills. The advice given adapts to the background of the individual, while at the same time attempting to motivate the client to want to analyse their data well.
  7. Consultant is a model: The consultant models clear thinking about research, clarifying that the role of research is to learn about the world.