def create(name, fidl, filterf, nod=True):
"""Use a fild file to create a metadata label table, using the json
filter file to convert, rename, and reorganize labels as needed.
Parameters
----------
name : str
The name of the csv file (inculde extension)
fidl : str
The name of the .fidl file
filterf : str
The name of the filer json file (see Note for format details)
nod : Boolean
Write SPM compatile NOD (Names, Onsets, Durations) .mat files
Notes
-----
Overall filterf has two goals. 1. Configure create() for the
current exp. 2. Remap/join labels/conditions from the fidl file
before using them in the csv. Fidl names are often rather
confusing and noisy.
The filterf should be valid json file, where the top level
is a list. The first element in that list must be the exp parameters
in a dict. The next and remaining elements are dicts whose keys
will become a colname in the csv. The values are the
fuzzy_label : label pairs.
Nothing is done to validate the filterf. Use with care and
double check the results.
For example (delete comments (#) before use/validation):
[
# The parameters
{
"triallen" : 5,
"condcol" : 2,
"trialcol" : 4,
"final_ncol" : 8
},
{
# The first filteration, col with be trial
# and each fidl match on the right gets
# coded as 'trial'
"trial": {
"1FaceCor1" : "trial",
"2FaceCor2" : "trial",
"3FaceCor3" : "trial",
"4FaceCor4" : "trial",
"5FaceCor5" : "trial",
"6HouseCor1" : "trial",
"7HouseCor2" : "trial",
"8HouseCor3" : "trial",
"9HouseCor4" : "trial",
"10HouseCor5" : "trial",
"11NoiseResp1" : "trial",
"12NoiseResp2" : "trial",
"13NoiseResp3" : "trial",
"14NoiseResp4" : "trial",
"15NoiseResp5" : "trial",
"16NoiseResp5" : "trial",
"16MultiResp" : "trial",
"17NoResp" : "trial"
}
},
}
# Col is rt, and both Cor1 and Resp1
# get remapped to rt1 (and repeat for 2, 3 ...)
"rt": {
"Cor1" : "rt1",
"Cor2" : "rt2",
"Cor3" : "rt3",
"Cor4" : "rt4",
"Resp1" : "rt1",
"Resp2" : "rt2",
"Resp3" : "rt3",
"Resp4" : "rt4"
}
},
{
# And so on....
"exp": {
"Face" : "face",
"House" : "house",
"Noise" : "noise"
}
}
]
"""
fidl_to_csv(fidlfile, name, 0)
## Write a csv to disk
filterdata = load(open(filterf, 'r'))
expdata = filterdata.pop(0)
condcol = expdata["condcol"]
trialcol = expdata["trialcol"]
triallen = expdata["triallen"]
for fd in filterdata:
fuzzy_label(name, condcol, _get_filt(fd), _get_name(fd), header=True)
# Add the name of the fidl file as a factor
dname = os.path.splitext(os.path.basename(fidl))[0]
if len(dname) < 3:
dname += "__"
sublab = {"trial": dname}
fuzzy_label(name, trialcol, sublab, "scode", header=True)
# Fill in TR time
tdur = {"trial": triallen}
tr_time(name, trialcol, tdur, drop=True, header=True)
ncol = expdata["final_ncol"]
fill_tr_gaps(os.path.join(os.path.dirname(name),
"trtime_" + os.path.basename(name)),
ncol,
fill='nan')
# Be SPM compatible?
if nod:
csvdata = pd.read_csv(name)
names = csvdata["rt"]
trials = csvdata["trial"]
onsets = csvdata["TR"]
durations = np.array([
triallen,
] * len(trials))
nod_mat(
names, onsets, durations,
os.path.join(
os.path.dirname(name),
"nod_" + os.path.splitext(os.path.basename(name))[0] + ".mat"))
"5FaceCor5": "trial",
"6HouseCor1": "trial",
"7HouseCor2": "trial",
"8HouseCor3": "trial",
"9HouseCor4": "trial",
"10HouseCor5": "trial",
"11NoiseResp1": "trial",
"12NoiseResp2": "trial",
"13NoiseResp3": "trial",
"14NoiseResp4": "trial",
"15NoiseResp5": "trial",
"16NoiseResp5": "trial",
"16MultiResp": "trial",
"17NoResp": "trial"
}
fuzzy_label(csv_f, condcol, triallab, "trial", header=True)
# ----
# 4. Add labels for rt independent of stim type
rtlab = {
"Cor1": "rt1",
"Cor2": "rt2",
"Cor3": "rt3",
"Cor4": "rt4",
"Resp1": "rt1",
"Resp2": "rt2",
"Resp3": "rt3",
"Resp4": "rt4"
}
## Dropping RT5 as it is missing form some
## SS and has low counts overall anyway
condcol = 2 ## The location in csv_f of the cond names
trialcol = 4 ## The location in csv_f of the trial desginator
# and convert the fidl to csv,
# and write that csv to disk.
csv_f = basename + ".csv"
fidl_to_csv(fidlfile, csv_f, 0)
# ----
# 2. Separate trials from events we want to ignore
triallab = {
"1Right" : "trial",
"2Left" : "trial",
"0fix" : "trial",
"4NoResp" : "trial"}
fuzzy_label(csv_f, condcol, triallab, "trial", header=True)
# ----
# 4. Add labels for rt independent of stim type
resplab = {
"Right" : "right",
"Left" : "left"}
fuzzy_label(csv_f, condcol, resplab, "resp", header=True)
# ----
# 5. Add labels for subject/fidl
sublab = {"trial" : basename}
fuzzy_label(csv_f, trialcol, sublab, "scode", header=True)
# ----
# 6. Expand labels so they cover every TR
def create(name, fidl, filterf, nod=True):
"""Use a fild file to create a metadata label table, using the json
filter file to convert, rename, and reorganize labels as needed.
Parameters
----------
name : str
The name of the csv file (inculde extension)
fidl : str
The name of the .fidl file
filterf : str
The name of the filer json file (see Note for format details)
nod : Boolean
Write SPM compatile NOD (Names, Onsets, Durations) .mat files
Notes
-----
Overall filterf has two goals. 1. Configure create() for the
current exp. 2. Remap/join labels/conditions from the fidl file
before using them in the csv. Fidl names are often rather
confusing and noisy.
The filterf should be valid json file, where the top level
is a list. The first element in that list must be the exp parameters
in a dict. The next and remaining elements are dicts whose keys
will become a colname in the csv. The values are the
fuzzy_label : label pairs.
Nothing is done to validate the filterf. Use with care and
double check the results.
For example (delete comments (#) before use/validation):
[
# The parameters
{
"triallen" : 5,
"condcol" : 2,
"trialcol" : 4,
"final_ncol" : 8
},
{
# The first filteration, col with be trial
# and each fidl match on the right gets
# coded as 'trial'
"trial": {
"1FaceCor1" : "trial",
"2FaceCor2" : "trial",
"3FaceCor3" : "trial",
"4FaceCor4" : "trial",
"5FaceCor5" : "trial",
"6HouseCor1" : "trial",
"7HouseCor2" : "trial",
"8HouseCor3" : "trial",
"9HouseCor4" : "trial",
"10HouseCor5" : "trial",
"11NoiseResp1" : "trial",
"12NoiseResp2" : "trial",
"13NoiseResp3" : "trial",
"14NoiseResp4" : "trial",
"15NoiseResp5" : "trial",
"16NoiseResp5" : "trial",
"16MultiResp" : "trial",
"17NoResp" : "trial"
}
},
}
# Col is rt, and both Cor1 and Resp1
# get remapped to rt1 (and repeat for 2, 3 ...)
"rt": {
"Cor1" : "rt1",
"Cor2" : "rt2",
"Cor3" : "rt3",
"Cor4" : "rt4",
"Resp1" : "rt1",
"Resp2" : "rt2",
"Resp3" : "rt3",
"Resp4" : "rt4"
}
},
{
# And so on....
"exp": {
"Face" : "face",
"House" : "house",
"Noise" : "noise"
}
}
]
"""
fidl_to_csv(fidlfile, name, 0)
## Write a csv to disk
filterdata = load(open(filterf, "r"))
expdata = filterdata.pop(0)
condcol = expdata["condcol"]
trialcol = expdata["trialcol"]
triallen = expdata["triallen"]
for fd in filterdata:
fuzzy_label(name, condcol, _get_filt(fd), _get_name(fd), header=True)
# Add the name of the fidl file as a factor
dname = os.path.splitext(os.path.basename(fidl))[0]
if len(dname) < 3:
dname += "__"
sublab = {"trial": dname}
fuzzy_label(name, trialcol, sublab, "scode", header=True)
# Fill in TR time
tdur = {"trial": triallen}
tr_time(name, trialcol, tdur, drop=True, header=True)
ncol = expdata["final_ncol"]
fill_tr_gaps(os.path.join(os.path.dirname(name), "trtime_" + os.path.basename(name)), ncol, fill="nan")
# Be SPM compatible?
if nod:
csvdata = pd.read_csv(name)
names = csvdata["rt"]
trials = csvdata["trial"]
onsets = csvdata["TR"]
durations = np.array([triallen] * len(trials))
nod_mat(
names,
onsets,
durations,
os.path.join(os.path.dirname(name), "nod_" + os.path.splitext(os.path.basename(name))[0] + ".mat"),
)
trialcol = 4 ## The location in csv_f of the trial desginator
# and convert the fidl to csv,
# and write that csv to disk.
csv_f = basename + ".csv"
fidl_to_csv(fidlfile, csv_f, 0)
# ----
# 2. Separate trials from events we want to ignore
triallab = {
"1Right": "trial",
"2Left": "trial",
"0fix": "trial",
"4NoResp": "trial"
}
fuzzy_label(csv_f, condcol, triallab, "trial", header=True)
# ----
# 4. Add labels for rt independent of stim type
resplab = {"Right": "right", "Left": "left"}
fuzzy_label(csv_f, condcol, resplab, "resp", header=True)
# ----
# 5. Add labels for subject/fidl
sublab = {"trial": basename}
fuzzy_label(csv_f, trialcol, sublab, "scode", header=True)
# ----
# 6. Expand labels so they cover every TR
final_ncol = 8
tdur = {"trial": triallen}
"4FaceCor4" : "trial",
"5FaceCor5" : "trial",
"6HouseCor1" : "trial",
"7HouseCor2" : "trial",
"8HouseCor3" : "trial",
"9HouseCor4" : "trial",
"10HouseCor5" : "trial",
"11NoiseResp1" : "trial",
"12NoiseResp2" : "trial",
"13NoiseResp3" : "trial",
"14NoiseResp4" : "trial",
"15NoiseResp5" : "trial",
"16NoiseResp5" : "trial",
"16MultiResp" : "trial",
"17NoResp": "trial"}
fuzzy_label(csv_f, condcol, triallab, "trial", header=True)
# ----
# 4. Add labels for rt independent of stim type
rtlab = {
"Cor1" : "rt1",
"Cor2" : "rt2",
"Cor3" : "rt3",
"Cor4" : "rt4",
"Resp1" : "rt1",
"Resp2" : "rt2",
"Resp3" : "rt3",
"Resp4" : "rt4"}
## Dropping RT5 as it is missing form some
## SS and has low counts overall anyway
fuzzy_label(csv_f, condcol, rtlab, "rt", header=True)