It is a controversial point, but as a referee, I have often wanted more data than are intended to appear in the final article. Sometimes the raw data themselves may be necessary for a proper review to take place even if, because of pressure of space, they cannot appear in the final journal article. In most cases, this request for the data has been met with alacrity, since anything that helps publication is usually of interest to the author. In a very few instances, nothing more has been heard from the author(s), which leads to definite conclusions about the quality of the work. I have had one instance where the data were supplied quickly but, where although the regression analysis produced the same results as quoted, I could not reproduce the statistical significance of the findings. Again, after raising a query, nothing more was heard by me from the authors. The problem is that obtaining the raw data, especially if not supplied in a computer-readable form, require considerable resources for carrying out full analyses to detect genuine evidence of misconduct. Carelessness alone is not fraud.
Often, no raw data are available and the scrutiniser must examine the text, tables, and graphs that are submitted. Experience and knowledge of the subject area may then mean that a whiff of suspicion arises on reading a manuscript. Various things can be done in checking the immediately available material. Some points that are indicators of problems are:
• Numbers that do not add up across a table.
• Graphs with different numbers of observations from those quoted in the text.
• P values that are quoted without the data necessary to estimate them.
Was this article helpful?