def projectionAnalyzer(ds, dpathUE="/UE", dpathECE='/evtcond', dpathFilters='/pca', bins=60, smooth=0, newpath='/hists'): '''construct the de-biased distribution of conditioned projections onto a filter or set of filters. The filters must have the same length as the conditioned events in dpathConditional. The distributions are calculated as histograms in bins of width "step"''' raw=ds.getSubData(dpathUE).getData() cond = ds.getSubData(dpathECE).getData() comps=ds.getSubData(dpathFilters).getData() hists=zeros((bins, comps.shape[1]), float32) ranges=[] for v in range(comps.shape[1]): filt=comps[:,v] cp=dot(filt, cond) rp=dot(filt, raw) st=min([rp.min(), cp.min()]) top=max([cp.max(), rp.max()]) bw=(top-st)/float(bins) ranges.append((st, bw)) rp=hist2(rp, bw, st, bins).astype(float32)/raw.shape[1] cp=hist2(cp, bw, st, bins).astype(float32)/cond.shape[1] z = arange(rp.shape[0]) if smooth: rp = dimr._lstsqO4(rp, z)[1] cp = dimr._lstsqO4(cp, z)[1] mask=nonzero1d(rp==0) put(rp, mask, 1) put(cp, mask, 0.0) hists[:,v]=cp/rp h=ds.header() h['Labels']=ds.getSubData(dpathFilters).getLabels() h['BinWidths']=[q[1] for q in ranges] h['StartBins']=[q[0] for q in ranges] ds.createSubData(newpath, data=hists, head=h, delete=True) return ds
def fitTSO4(ds, newpath=""):
chans = []
h = ds.header()
fs = h["SamplesPerSecond"]
start = h.get("StartTime", 0)
for i in range(ds.data.shape[1]):
dat = ds.data[:, i]
v = DR._lstsqO4(dat, fs, start)
chans.append(v[1])
if newpath:
ds.createSubData(newpath, column_stack(chans), h, True)
else:
echans = []
for i in range(ds.data.shape[1]):
echans.extend([ds.data[:, i], chans[i]])
ds.datinit(column_stack(echans), {"SampleType": "ensemble", "Reps": 2, "SamplesPerSecond": fs}, True)
def fitFunctionsO4(ds):
dats = ds.getElements("Data", {"SampleType": "function"})
mfv = max([d.data[:, 0].max() for d in dats])
vals = array([e.attrib("length") for e in dats])
order = vals.argsort()
off = 0
for i in order:
f = dats[i]
print i, f.attrib("length"), f.name()
h = f.header()
x = f.data[:, 0]
y = f.data[:, 1]
pars, vals = DR._lstsqO4(y, x)
f.data[:, 0] += off
off += mfv
x = f.data[:, 0]
h.update(dict(zip(["m", "gamp", "mean", "std", "e"], pars)))
d = column_stack([x, vals])
h["style"] = "line"
path = f.dpath() + "_fit"
ds.createSubData(path, d, h, True)