888 residuals #939
888 residuals #939
Conversation
…ts of residuals across all simulations by output
TODO: Output and SimulationSets needs to manage the property
|
some tests failed on all systems... |
NAMESPACE
Outdated
| @@ -1,5 +1,7 @@ | |||
| # Generated by roxygen2: do not edit by hand | |||
|
|
|||
| export("geomean*sd") | |||
| export("geomean/sd") | |||
There was a problem hiding this comment.
Is this a valid function name
There was a problem hiding this comment.
@pchelle Please rename the functions. You have to call the functions in quotes otherwise
`geomean*sd`(c(1,2,3))
Not sure if export must contain the quotes as well then, but having such a function name is pretty crazy from my point of view.
There was a problem hiding this comment.
I agree. I don't even know why this is allowed
(hopefully, this will enforce its installation in the vignettes)
Checking if rlang might be responsible for errors in tests
Codecov ReportBase: 72.59% // Head: 70.47% // Decreases project coverage by
Additional details and impacted files@@ Coverage Diff @@
## develop #939 +/- ##
===========================================
- Coverage 72.59% 70.47% -2.13%
===========================================
Files 58 58
Lines 7379 7719 +340
===========================================
+ Hits 5357 5440 +83
- Misses 2022 2279 +257
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
| ymax = tlf::tlfStatFunctions$`Percentile97.5%`, | ||
| yCaption = "median", | ||
| # The unicode characters below are superscript th | ||
| rangeCaption = "[2.5\u1d57\u02b0-97.5\u1d57\u02b0] percentiles" |
There was a problem hiding this comment.
when entering this in R I get the following result:
> "[2.5\u1d57\u02b0-97.5\u1d57\u02b0] percentiles"
[1] "[2.5\u1d57ʰ-97.5\u1d57ʰ] percentiles"
But the unicode seems to be correct.
Does it look good when creating a report?
There was a problem hiding this comment.
On my local computer it does.
I checked the default printing documentation that indicates the following:
Whether a character is printable depends on the current locale and the operating system
There was a problem hiding this comment.
And when checking using vscode on the md document, it also does look good.
My plots also looks okay, but I don't know if they will on your side
There was a problem hiding this comment.
interesting...
OK, let's keep it for now AS IS, I will check later how the reports look like e.g. on my machine
R/reportingengine-env.R
Outdated
| ymax = "geomeanMultipliedBySD", | ||
| yCaption = "geometric mean", | ||
| # The unicode character below is supposed to be */ symbol | ||
| rangeCaption = "mean \u22c7 geometric SD range" |
There was a problem hiding this comment.
was the caption supposed to be "mean */ geometric SD range"?
because when entering this in R I get:
> "mean \u22c7 geometric SD range"
[1] "mean ⋇ geometric SD range"
There was a problem hiding this comment.
Yes, I checked for the unicode of +/- and found there was also one for */
If it looks weird I can revert and use */ instead
There was a problem hiding this comment.
Let's use */.
After looking at the symbol in very big resolution (https://www.compart.com/en/unicode/U+22c7) I finally understood what it is, but in the smaller size it just looks like "*"
| simulatedData <- removeNegativeValues(simulatedData, dataMapping$y) | ||
| observedData <- removeNegativeValues(observedData, dataMapping$y) | ||
| lloqData <- removeNegativeValues(lloqData, dataMapping$y) |
There was a problem hiding this comment.
There are some problems with this solution.
But I would suggest to address them in the separate PR in order not to block this PR for even longer.
(just created an issue for this: #942)
I have already created the issue for this in the ospuite-R (because residuals are calculated there as well):
Open-Systems-Pharmacology/OSPSuite-R#1091
In short:
- Removing <=0 values (especially observed values) is not correct. E.g. if we have
observed = 0andsimulated >>0, we just ignore a significant residual in such a case. - Very small positive values are problematic for log residuals as well. E.g. having
observed = 1e-100andsimulated = 1e-200would produce a huge log residual, but in fact the difference is irrelevant - both values are numerical zero.
For this, PK-Sim/MoBi introduces the LogSafe function, which is used for calculating log residuals and defined as:
LogSafe(Y) = Log(Max(Y, LogSafeEpsilon)) (with LogSafeEpsilon currently defined as 1E-20)
And Lin residuals are calculated with all the values AS IS
No description provided.