Physicians and p-values – like children and ketchup.
A fundamental part of an experiment or trial is its design. The definition of the main question, the outcome, and the procedures used to obtain the respective measurements, are essential for the success of a study. A flaw in the design might lead to bias, reduced validity, or failure of the study.
The statistician plays a vital part in the planning phase of a study. He or she is well versed in different designs and can advise the researcher on most issues related to a scientific project.
Developing a study, devising the statistical plan, and writing a sound study protocol will take time. The benefit, however, is that hasty ad hoc decisions due to unexpected contingencies during the study are avoided. Ambiguous or conflicting results and the extended analysis of abundant variables are eliminated. The writing of the final report is facilitated due to proper planning, a clear outcome, and well-defined conclusions.
Contact me at an early stage of your research project, preferably before you submit your proposal to any committees for approval or funding. I can support you in writing a cogent proposition while pointing out potential shortcomings, such as
- variables not measured one can not control for later,
- single instead of repeated measurements prohibiting the assessment of variance components or interactions,
- missing observations at critical stages in a longitudinal study,
- recording the sequence of collected samples but not the actual time,
- ignoring competing risks, e.g. define a cardiac event or death both as main outcome,
- defining groups implicitly by a time-dependent variable,
- introducing ties by imprecise measurements,
- not measuring the time to an event of interest,
- insufficient events in a study with a binary outcome,
- assessing reliabilities with a correlation coefficient.
I provide a comprehensive and well documented report, briefly describing the received data set, giving summary statistics of the variables of interest, exploring their relationship graphically, and checking the data for plausibility.
In case of further analysis, I summarize the results obtained in the form of a continuous text interspersed with tables and plots, if necessary explaining the rationale of the selected methods. This report is either in English or German. All plots are supplied as pdf files. If a publication of the results is intended, a “Methods Section” is written that can be used for “copy/paste.”
The data analysis is mainly performed in R, a programming language for statistical computing. Every step, from the read-in and initial processing of the data to the final analysis, is documented and reproducible.
Later changes in the data set are easily implemented. If a subject has to be excluded or the analysis should be based on new observations, the code is adjusted and simply re-compiled.
Often, a researcher has a good working knowledge in statistics. He or she is familiar with SPSS or GraphPad Prism, and can apply the most common methods. However, when the experiment and the data structure become more involved, advice and a few hints are needed to obtain reliable results or to avoid pitfalls.
I offer statistical consulting with respect to study design, structuring of the data, recommended statistical methods, or technical issues. It is usually helpful to have a study plan or to see the data set. In case of a telephone conference, questions provided in advance are highly appreciated.
To write a scientific report is a big challenge. Being restricted to a limited number of pages, months of research, references, analyses and results should be presented in an intelligible manner, in comprehensible order, and in a well balanced depth of details.
I offer thorough proofreading of manuscripts, albeit with respect to quantitative results, their formal presentation, the coherence of thoughts, or language issues.