My guess is that you want to estimate the shape parameter and the scale of the Weibull distribution while keeping the location fixed. Fixing loc assumes that the values of your data and of the distribution are positive with lower bound at zero.
floc=0 keeps the location fixed at zero, f0=1 keeps the first shape parameter of the exponential weibull fixed at one.
>>> stats.exponweib.fit(data, floc=0, f0=1)
[1, 1.8553346917584836, 0, 6.8820748596850905]
>>> stats.weibull_min.fit(data, floc=0)
[1.8553346917584836, 0, 6.8820748596850549]
The fit compared to the histogram looks ok, but not very good. The parameter estimates are a bit higher than the ones you mention are from R and matlab.
Update
The closest I can get to the plot that is now available is with unrestricted fit, but using starting values. The plot is still less peaked. Note values in fit that don’t have an f in front are used as starting values.
>>> from scipy import stats
>>> import matplotlib.pyplot as plt
>>> plt.plot(data, stats.exponweib.pdf(data, *stats.exponweib.fit(data, 1, 1, scale=02, loc=0)))
>>> _ = plt.hist(data, bins=np.linspace(0, 16, 33), normed=True, alpha=0.5);
>>> plt.show()
