How can a statistical consultant improve your research design?

A statistical consultant, is a qualified statistician with experience who provides guidances to researchers. A good consultant will make sure that he/she understand the questions to address as well as the underlying context. The consultant will make sure that a solid experimental design is developed as well as a statistical analysis plan (before the data are collected). He/she will highlight the scope and limitations (if any) of the results of the statistical analysis/modelling.

Statistical consultant should assist in determining sample sizes; optimising experimental designs and selecting appropriate statistics. For instance, they advise on survey strategies and experiment designs to secure meaningful findings. The consultant should be involved right at the start of the research, communication of goals must be made clear, and they have to be considered during any choice-making relating to design. This collaboration makes sure robust and meaningful statistical analysis.

A statistical consultant will make sure that the research/business question is clearly formulated and can be translated into statistical terms. The consultant will also help ensure that the collected data are useful and precise enough to address the research/business question. The consultant will provide guidance on the choice of the experimental design and the analysis plan. He/she will provide expert guidance to minimise the risk of bias in the results, highlight the scope and potential limitations in the findings.

Securing a statistical consultant involves diverse approaches based on your needs and resources. Universities often provide consulting services through dedicated centers. Professional statisticians' associations offer directories of consultants, and online platforms like Upwork connect researchers with qualified professionals. Posting project details on these platforms allows you to review proposals, profiles, and communicate directly with potential consultants.

Whenever looking for a statistical consultant, make sure that you look for someone with a degree in statistics so that you know that they master the foundations and the various statistical tools & modelling techniques. Try to find statisticians who are at the forefront of new methodological advances. One should also look for experienced consultants. Make sure that you explore their accomplishments and the feedback that they got from past & current clients. See how rigorous and creative they can be.

With over 30 years experience, I found that the most common mistakes that clients make: 1) they try formulating their questions into statistical terms. Please don't. Simply explain your problem. It's the consultant's job to translate it statistically. 2) Don't hide any information from the consultant. This will invariably impact the data collection process as well as the quality of the statistical analyses/modelling. 3) Whenever planning an experiment in the lab or in a plant, make sure that the consultant sees the manipulations/treatments involved. This is money well invested as it helps the consultant to understand the entire process (scope) and the potential sources of variation/bias.

If you have determined that you do require a statistical consultant then it is better to start sooner than later. It is because every step, from thinking of doing an experiment to actually doing it, is critical. How you design your experiment, what data to collect, how to collect, how many samples, replications, what analysis, what tools, etc. everything is going to impact your conclusion. And every step is an opportunity to make your study better or worse. And if you have already collected the data that is unrepresentative of your population then your consultant won't be able to help much. That being said, make sure your consultant is also qualified, because you don't want to spend a lot of money for wrong advice.

Some good practices for research design statistical consulting: - Involve the statistical consultant early. Once data are collected the consultant can only play at salvage in data analysis of data which has not been optimally structured for the desired analyses. - Statistical consultants are not one size fits all, find one with design of experiments (DOE) expertise preferably in a research data context similar to yours. - Expect how to solve the research problem efficiently to be different than whatever was "planned". -Expect the research focus to become better defined; a good statistical consultant operates in a problem solving context not just as a statistician. - Expect data accumulation in stages rather than all at once.