Time resolved analysis

[AlexanderNiedrig]

Added the timeResolvedAnalysis script in a subfolder together with an example for a directory (its probably non functional as most data was deleted to lower file size). Most important naming conventions should be in the config as well and will be in the manual.

[JohannaRahm]

With the given example data I get following error message. Could you please share a fraction of working data (not over github – please remove the uploaded example data folder)?

[GLOBAL_DIR]
# define the directories to the .h5 files. They MUST follow the naming scheme something_coverslip_something. They must have been created by filtering for every parameter
dir = D:\Chemie_phd\SPTAnalyser\code_reviews\SPTAnalyser\timeResolvedAnalysis\required_structure_example\+HBEGF+CTX\230627_filtered

[USE_TIMESTAMPS]
#whether or not to use timestamps instead of cell numbers for binning
use_timestamps = false

[CS_DIRS]
#directories of the analyzed coverslips
dir1 = D:\Chemie_phd\SPTAnalyser\code_reviews\SPTAnalyser\timeResolvedAnalysis\required_structure

Please give variables and functions lower case names, add spaces between “#” and comment, use “”” instead of ‘’’ as docstring symbol, add description to function names (partly only the function variables are described, but not the function itself), and correct typos (they are all over the place).

[JohannaRahm]

• Make script compatible with h5py versions > 3.0
• Rename main folder TR_Analysis → timeResolvedAnalysis

Config.ini

• Add general description to config.ini
• Stat settings
  • Have two separate alpha values for (1) normality test (2) significance test

Stats.h5

• General remarks
  • Add units to columns
  • Average over all coverslips and save somewhere in stats (e.g. “global” folder)
• Metadata
  • The information about how many cells are averaged is missing. In the test dataset use_timestamps is set to False, but no info about the number of cells is given

General remarks about tests (normality + significance)

• Rename folders with underscore and lowercase (confinement Radius → confinement_radius, Diffusion Coefficient → diffusion_coefficient, …). Please also apply this renaming scheme to the folders in TRplots
• Rename saved files, examples below
  • confinement → test_normality_confinement_radius
  • diffusion → test_normality_D_confined
  • fractions: Pfree → test_normality_percentage_free
  • number of segments: Nglobal → test_normality_n_segments_global
  • length of segments: Lglobal →test_normality_length_segments_global
  • length of segments: Lglobal →test_significance_length_segments_global
• Add column about number of cells per cell range (this helps to determine if NaN values in columns are due to low test statistics or errors)
• For the cell ranges of 21-25 an error occurs (for all tests) → NaN,NaN,Dataset too small,NaN,NaN,Dataset too small (about 6 cells should be in that range for CS4 and CS6)

Significance tests

• Use Wilcoxon and paired-t test instead of Mann Whitney U and Wilcoxon for significance tests. We are looking at conditioned variables (therefore paired tests) and depending on the results of the normality tests either Wilcoxon or paired t-test are of interest to the user.
• Compared bins should be 1-5 cells vs 6-10, 1-5 cells vs 11-15, … but are atm 1-2 vs 6-10, 1-3 vs 11-15, … there seems to be an error in the selection of the first comparison batch

TR_plots

• Vertical line for ligand addition seems to be incorrectly shifted to ~3 cells instead of 5 cells.

[JohannaRahm]

The PR looks good!