i28 generalise upper and lower detction limits #29
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #29 +/- ##
==========================================
+ Coverage 93.47% 94.53% +1.05%
==========================================
Files 9 9
Lines 506 567 +61
==========================================
+ Hits 473 536 +63
+ Misses 33 31 -2 ☔ View full report in Codecov by Sentry. 🚨 Try these New Features:
|
R/plot.R
Outdated
| geom_smooth(aes(x = time_since_last_exp, y = value, colour = titre_type)) + | ||
| geom_point(aes(x = time_since_last_exp, y = value, colour = titre_type, shape = censored), | ||
| size = 0.5, alpha = 0.5) + | ||
| geom_smooth(data = subset(data, !censored), |
There was a problem hiding this comment.
Added the censored indicator in to the input data plot as well, and excluding censored values from the spline - do we think this is the best way to deal with these data points? In the legacy data, the censored values are for some reason much lower than any other value in the dataset (so presumably lower than the actual assay detection limit), but if this wasn't the case would probably make sense to include them.
There was a problem hiding this comment.
Yeah I think so, might be worth changing the point marker to show a value is censored? Or add the lower and upper limits to the plot (just dashed lines?) I think should illustrate the point clearly
There was a problem hiding this comment.
yeah the shape is different if censored, but dashed lines could be good too 👍
| dt_out <- data.table::copy(dt_in) | ||
| for (var in vars_to_transform) { | ||
| if (simplify_limits == TRUE) { |
There was a problem hiding this comment.
At the moment am just warning if there are values outside of provided detection limits... but perhaps if there are, we should also keep this logic, i.e. set all higher values to the upper limit and all lower values to the lower limit?
There was a problem hiding this comment.
I think given the new warning this will work fine now right?
There was a problem hiding this comment.
Yeah, so we are now doing this when the data is first passed in, iff the strict_upper_limit/strict_lower_limit args are TRUE.
R/plot.R
Outdated
| geom_smooth(aes(x = time_since_last_exp, y = value, colour = titre_type)) + | ||
| geom_point(aes(x = time_since_last_exp, y = value, colour = titre_type, shape = censored), | ||
| size = 0.5, alpha = 0.5) + | ||
| geom_smooth(data = subset(data, !censored), |
There was a problem hiding this comment.
Yeah I think so, might be worth changing the point marker to show a value is censored? Or add the lower and upper limits to the plot (just dashed lines?) I think should illustrate the point clearly
There was a problem hiding this comment.
These are the plots I'm thinking of having the upper and lower limits on
There was a problem hiding this comment.
These look different because as well as adding upper/lower limits to the plot, I've made the points smaller and added a tmax argument with a default of 150 days post exposure
There was a problem hiding this comment.
I'm still including censored points in the smoothing function, I thought that was probably the expected behaviour, but the artifically low values in the legacy data (below the real titre detection limit) does distort them.
hillalex
force-pushed
the
i28
branch
2 times, most recently
from
aa2cc63 to
27446ca
Compare
| values_above <- max_value > private$upper_censoring_limit | ||
| values_below <- min_value < private$lower_censoring_limit | ||
| if (strict_upper_limit) { | ||
| if (values_above) { |
| dt_out <- data.table::copy(dt_in) | ||
| for (var in vars_to_transform) { | ||
| if (simplify_limits == TRUE) { |
There was a problem hiding this comment.
I think given the new warning this will work fine now right?
Adds
upper_detection_limitandlower_detection_limitarguments to thebiokineticsclass constructor, and takes the largest/smallest data values as defaults. These are used when scaling the data (if providing natural scale data), to determine which points are censored, and to centre the likelihoods for upper/lower censored data points.Also adds
truncate_upperandtruncate_loweras boolean flags - if TRUE, this sets all points above/below the limits to those limits. Points at the limits are treated as censored.