Given a few examples of experiments over a time (or spatial) course,
'NITPicker' selects a subset of points to sample in follow-up experiments,
which would (i) best distinguish between the experimental conditions and the
control condition (ii) best distinguish between two models of how the
experimental condition might differ from the control (iii) a combination of
the two. Ezer and Keir (2018)
This contains the code for generating all the figures in 'Selection of time points for costly experiments: a comparison between human intuition and computer-aided experimental design.
in Pathfinder.R: drawSurveyBasicAnalysisFigures: figures related to the survey testCanada: figures related to the Canada dataset testBerkeley: figures related to the Berkeley dataset
27/06/18: This new version has a few major changes. Firstly, it was noticed that the 'fast mode' sometimes gave drastically different results compared to the 'un-fast mode'. This has been fixed and now the fast mode produces much more stable results. Secondly, we added a mode that allows the user to run NITPicker on multiple genes at once.
22/04/18: This version has a 'fast mode' which switches the order of the summation and integration in the f1, f2, and f3 estimates, which can improve the speed of the algorithm. In addition, the previous version of f3 has the L-4 estimate instead of the L-2, but this bug has now been fixed.
19/04/18: This version of the tool provides methods for identifying the best subset of time points to test under the f1, f2, and f3 metrics described in the vignettes. Splines are used to interpolate between points prior to sampling from the probability distribution of functions, but it is currently not used to interpolate between points when calculating the L2 error as part of the dynamic programming algorithm in order to decrease the runtime.