The model chosen allowed the effect of the modelled-PCM concentrations to vary across space which showed a slight improvement over the model that assumed the effect was constant (RMSPE of 0.255 compared to 0.258).
Each of the models described above assume the effect of the modelled-PCM concentrations is constant across space.
Even though the bias in Model 9 (0.013) is slightly higher compared to Model 8 (0.009), Model 9 is a better model compared to Model 6 as it treats the effect of the modelled-PCM concentrations constant across space as the RMSPE decreases by around 1% and the coverage probability is closer to the nominal value of 95%.
Therefore, to allow flexibility in the effect of the modelled-PCM concentrations to vary across space, Models 8 and 9 contain an interaction term between the log modelled variable and the easting and northing coordinates of the location of the automatic monitors and diffusion tubes, given as log modelled:easting and log modelled:northing.
We also considered allowing the effect of the modelled-PCM concentrations to vary as a quadratic surface in location added into Model 9 as log modelled:easting + log modelled:northing, but this did not improve the results.
We proceed to discuss the territorial cohesion effects of the modelled accessibility changes.
The experimental strategy was to study this 3D cell aggregation in terms of the cell phenotypes following short-term culture (up to 48 h, as a time that allows the formation of multicellular aggregates) and long-term (up to 2 weeks) culture, to analyse the effects of this modelled microgravity system on cell behaviour.
This model essentially assumed no further effect of the intervention but modelled out fully the possible QALY gains from the medium-term (11 years) differences in mortality and event rates.
Cubic regression splines may also be used to model and to test non-linear effects of covariates on event-specific hazard functions in a simple way: since any model with a linear effect of a given covariate is nested within models with a non-linear effect of the same covariate modelled using regression splines, a simple LR test may be used.
We adopted standard methods for interrupted time-series to assess the effects of the smoking ban [19]: the immediate effect of the ban was modelled as a step function, including an indicator variable that changed in January 2005, while the gradual effects were studied with an interaction term between the effect of the ban and time.
Web figure 2 shows the effect of all modelled uncertainties on the primary output of incremental cost per QALY associated with the screening programme, and web figure 3 shows the distribution of the other outputs of the model.
