Statistical Analysis
The statistical analysis that was done for this pilot study can be split
up in four parts.
The first step was aligning the data pertaining to movement with the
data regarding urine production. The first hurdle that had to be crossed
was that of the discrepancy in data collection moments for both urine
production as well as that of movement. As mentioned above, participants
were asked to collect eight urine samples. Contrary to the urine
sampling, participants were asked to fill in hourly updates regarding
their majority movement of the past hour, resulting in 24 inputs of data
for the same day regarding movement. The solution to this was clubbing
three inputs of data of movement. Here the range of movements no longer
went from 0-4, but now went from 0-12. Because of this, there was now a
symmetry in the data input of urine samples as well as that of movement.
The second step was to change the database format in its entirety. The
database was in a wide format. This had to be changed to a long format,
as otherwise, the analysis that had to be done would have been
statistically challenging. As such every participant in the study had 16
different data entries each pertaining to 1 of 8 timeslots of both days.
The third step was to test the hypothesis by conducting a basic
statistical test such as the Kruskal Wallis and the Mann Whitney U test.
To successfully conduct these tests, movement was divided into three
categories, defined as 1,2 and 3. They respectively stood for less,
moderate and more movement and physical activity. A p-value of
<0,05 was considered as statistically significant.
The fourth and final step was the main statistical analysis. Mixed model
statistics were used to analyze the impact of movement on urine
production. Here diuresis, osmolality, creatinine and sodium were
compared with movement to find any form of association. Parameters of
fixed effects were estimated using maximum likelihood estimation and
reported as standardized regression coefficients (β) with their
respective standard error. A p-value <0.05 was considered as
statistically significant. These tests were conducted for the entire day
as well as day and night separately. Factors such as “BMI” and “age”
were defined as confounders and included in the mixed model.
Statistical analysis was carried out using SPSS 24.