# passive properties: we weren't actually using the stimulus protocol. it seems only useful for testing a TEA case to see that it is similar to control
# in Ransdell 2013 they calculated Rin by: Rin was measured in TECC with 10 negative current injection steps (t⫽ 2s, ⫺1to ⫺10 nA).. This is not what I do
# could not find a reference for tau, so I didn't use it. not sure how to calculate it anyway
# https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6328273/ shows there is feedback between LC and SC?
# should the synaptic input increase in magnitude not just frequency? This will require >1 Exp2Syn
# in LV2, should the siz gleak be based on the passing soma leak, or pick a new one, or pick the same one that was randomly picked by the neurite?
#https://www.neuron.yale.edu/phpBB/viewtopic.php?t=2991 suggests that this would be faster if the templates were made in hoc instead of python classes
#unclear why lv2 is slow. using a vecstim vs iclamp seems to make no difference. It is probably just that the cell is larger and has more channels?
#parallelContext has not been tried recently to speed things up
#calcium infity pool is not set in the refactored code. Should I set it? to what value?
#make the directory initialization more generic so we don't have to write the path each time
#like make a constant function be the path, and the lists be constants, then set them outside the directory
#so they are not included in pushing and pulling to git.
#printNetVoltages is probably better done by using the unpacking feature of plt.subplots instead of the clumsy array list?
#generalize the pdf printing
#add a function that returns the parameter number and prints it 
# the pdf resulting from two uniform pdfs multiplied is normal. the passing gmax distributions are mostly uniform,
#and the resulting distribution of the summary statistics (area, spb,etc) is gaussian. Should a linear combination of the conductances predict
# the summary statistic?
# different TEA modifiers should be tried, like reduce by 90% not by 97%
#try rerunning LV2 with the updated eventtimes function
#use an adjacency matrix or something for LV3
#add something for visualizing currents
#make raster plot into a function
#TEA responses almost all have spikes after the large burst, this is not very like what is in the literature
# plot conductance labels on x axis, and LC number on the y axis, and each dot is a standardized parameter
# plot the electrophysiological feature value vs the parameter as it is changed from its min to max, then fit a polynomial to it. the derivative poly is a measure of the model's sensitivity to changes in that parameter.
# this can also be done for the network, see https://www.jneurosci.org/content/29/17/5573



#########VALIDATTION CHECKS##############
#rerun LV2 so without an active neurite (gleak only) so we can see how many pass lv2 without it
#a validation check for the use of the restricted range. rerun everything and show that (probably) next to no cells pass lv2 or 3. ideally, you rerun for both lv1restricted and lv2restricted to test
# rerun using stimulus protocol to see if/how the results differ. This is not necessary since we weren't using it to begin with, but it is a good comparison with Lane's work

#####
frequency gets rounded in making the event times, so find a way to use this instead of the scfrequencies which make it.


### Q's
https://www.jneurosci.org/content/29/17/5573
mention that "For instance, having subunits of multiple channel types under the control of a common transcription factor might be simpler than independent transcriptional control of each channel type. This could result in different subunits being expressed in fixed ratios. Thus, the presence of correlations between conductances does not automatically imply that these correlations are required for proper electrophysiological behavior.'
Do we have a way to show which interpretation would be more consistent with model and lab data?


the completely unknowns:
gSIZ
gSyn
gleak SIZ
gNa SIZ
gKd SIZ