There was generally a lack of consistency in the trends of variance for the three temporal scales at the smallest spatial scale of plots.
The trend of similar variances between spectral deconvolution and peak integration held for a range of experimental conditions (data not shown), such as changes in the gradients or nongradient selected versions of the experiment.
In the 633 patients followed up, we first examined variations of baseline demographics, risk factors, and comorbidities across the three severity groups, assessing significance by means of χ test for trend for categorical variables and trend test (analysis of variance with polynomial contrast) for continuous variables.
There was a change over occasions for both vignettes with beliefs becoming closer to those of professionals (significant linear trend in analysis of variance), but also a tendency for beliefs to revert somewhat at follow-up (as indicated by a significant quadratic trend).
However, social distance decreased over occasions for both groups (significant linear trend in analysis of variance), with some increase again at follow-up (significant quadratic trend).
To compare categorical variables between the three groups we used χ tests for trend, with analysis of variance and the Kruskal-Wallis tests used to compare normally and non-normally distributed data across the groups.
The potential trends in continuous variables across the testosterone quartiles were examined using the test for linear trend available in analysis of variance (ANOVA).
Number of patients achieving HbA1c level <7.0%, number of withdrawals at 1-year follow-up, and number of hypoglycemic episodes were tested using the Cochran-Armitage test or trend test using analysis of variance according to baseline HbA1c levels and duration of diabetes.
For ordinal predictor variables (such as interest in caring for immigrant patients), we obtained tests of linear trend within the analysis of variance (this test separates the between-group sum of squares into a linear component, using one degree of freedom, and an extra-linear component).
In Table 1, the annual age standardised rates for LC are reported by altitude, urbanisation, deprivation and pleural cancer mortality, with LC showing a linear trend with altitude (analysis of variance: P<0.001) and being significantly less frequent in the 2150 'rural' municipalities.
