As the investigators`s original question relates the occurrence of
hypotension over the entire dialysis period,
it makes most sense to include as much data as practical into the
initial descriptive analysis.
Boxplots (or alternatives such as Violin plots) are very informative
graphical methods. Since the distributions appear fairly symmetric,
the mean and standard deviations are appropriate descriptive
statistics. If the data is quite skewed, reporting medians and
quartiles is preferred.
Reporting the occurrence of hypotension highlights a technical question. One can report the occurrence rates per dialysis session, or on a per person basis. The latter could be described by giving the number of patients experiencing any hypotension (by period and temp) or by describing patient using the proportion of sessions where hypotension occured (by period and temp).
Similarly, since the eventually analysis will be based on mean blood pressures for patients within period and temperature, it is also meaningful to create boxplots for these aggregate data points.
Whatever approach is taken, it is very important to specify exactly how the variables represented in plots and tables relate to the original data provided
In consideration of the formal analysis using the Hill`s Armitage
approach, one should first provide summaries and plots of the
change scores that will be used in the analysis.
As well the approach to calculate change scores (mean of differences or
difference of means) needs to be specified. Most importantly,
report of the analysis should emphasize inferences concerning
the magnitude and direction of treatment and
time effects by providing confidence intervals.
Nearly everybody answered this last question in a pragmatic way, by simply conducting an analysis of covariance. Once again, it is very important to describe the variables analysed in terms of the original data, and make it clear whether the analysis was on means or individual values from visits.
Tables and Figures should have descriptive headings or labels and should be presented following (not before!) a description of contents.
Avoid ambigous references to Group 1 or 2 or A or B - use terminology specific to the data, e.g. Cold and Warm,