Collect various analyses plots

[Radonirinaunimi]

Given that we are more or less ready to run a full fit with our machine learning framework, it would be good to collect here various plots: data vs Yadism (in order to make sure that the coefficients are correct), kinematics, covariance matrices, etc.

[Radonirinaunimi]

Results below were generated using 1d40aa7:

---

Kinematics:

---

Combined covmat (Normalised):

---

BEBCWA59_F2 (Normalised):

BEBCWA59_F3 (Normalised):

CCFR_F2 (Normalised):

CCFR_F3 (Normalised):

CDHSW_DXDYNUB (Normalised):

CDHSW_DXDYNUU (Normalised):

CDHSW_F2 (Normalised):

CDHSW_F3 (Normalised):

CDHSW_FW (Normalised):

CHARM_F2 (Normalised):

CHARM_F3 (Normalised):

CHARM_QBAR (Normalised):

CHORUS_DXDYNUB (Normalised):

CHORUS_DXDYNUU (Normalised):

CHORUS_F2 (Normalised):

CHORUS_F3 (Normalised):

NUTEV_DXDYNUB (Normalised):

NUTEV_DXDYNUU (Normalised):

NUTEV_F2 (Normalised):

NUTEV_F3 (Normalised):

[juanrojochacon]

Thanks @Radonirinaunimi , the kinematic plot looks super nice! Maybe we can use log scale also for x axis? Clearly we have quite a bit of data points for low-Q, so we can count on a reliable extrapolation.

Question: do we impose right now any cut in W? If we want to restrict ourselves to inelastic scattering and avoid the resonance region we have to impose W > 1.8 GeV or what is equivalent W^2 > 3.25 GeV^2. How does the kin plot change with this restriction?

[Radonirinaunimi]

Thanks @Radonirinaunimi , the kinematic plot looks super nice! Maybe we can use log scale also for x axis? Clearly we have quite a bit of data points for low-Q, so we can count on a reliable extrapolation.

Yes, we do indeed have quite some points in the low-$Q^2$, so this is good news. Below is the plot with log scale in the $x$ axis.

Question: do we impose right now any cut in W? If we want to restrict ourselves to inelastic scattering and avoid the resonance region we have to impose W > 1.8 GeV or what is equivalent W^2 > 3.25 GeV^2. How does the kin plot change with this restriction?

For the time being we do not impose a cut on $W$, but we indeed definitely should. We will implement this cut and will post here the updated kinematic plot.

[alecandido]

Ah @Radonirinaunimi, you can also select cuts directly from the CLI, there is also the help to show how :)

(not yet on W, just for x and Q2 for the time being, but in order to cut on W we only need to compute it)

[Radonirinaunimi]

Ah @Radonirinaunimi, you can also select cuts directly from the CLI, there is also the help to show how :)

But this is only for $(x, Q^2)$.

(not yet on W, just for x and Q2 for the time being, but the in order to cut on W we only need to compute it)

Ah, yes! You mentioned it already :)

[Radonirinaunimi]

Here are the kinematic plots (#23, 84d131f) with a $W$ cut ($W^2 \geq 3.5~\mathrm{GeV}^2$):

Linear $x$

Log $x$

As you can see, the cut cuts off a few of the low-$Q^2$ datapoints from the BEBCWA59 experiment.

[alecandido]

@Radonirinaunimi if it is not too difficult, it would be interesting to plot a line in this space, corresponding to the $W^2$ cut, i.e. the function:

$$ Q^2_{min}(x; W^2_{cut}) $$

Just for visualization.

[Radonirinaunimi]

@Radonirinaunimi if it is not too difficult, it would be interesting to plot a line in this space, corresponding to the W2 cut, i.e. the function:

Qmin2(x;Wcut2)

Just for visualization.

This is very interesting indeed and actually is already available:

nnusf/src/nnusf/plot/kinematics.py

Lines 62 to 66 in b3c9bec

	if wcut:
	min_value, max_value = ax.get_xlim()
	xvalue = np.arange(min_value, max_value, 5e-2)
	fq2 = lambda x: x * (3.5 - 0.95) / (1 - x)
	ax.plot(xvalue, fq2(xvalue), ls="dashed", lw=2)

For the time being, the $W^2_{\rm cut}$ is hard-coded to some value, I will modify it such that we can have it as an input argument.

[alecandido]

For the time being, $W^2_{\rm cut}$ is hard-coded to some value, I will modify it such that we can have it as an input argument.

Good, the easiest upgrade you can do is to turn wcut from a boolean to a number, this way:

def plot(
    groups: dict[str, list[list[float]]],
    wcut: Optional[float] = None,
    xlog: bool = True,
    ylog: bool = True,
) -> matplotlib.figure.Figure:

(Optional you find inside typing module). And of course checking for it accordingly:

 if wcut is not None:
     min_value, max_value = ax.get_xlim() 
     xvalue = np.arange(min_value, max_value, 5e-2) 
     fq2 = lambda x: x * (wcut - 0.95) / (1 - x) 
     ax.plot(xvalue, fq2(xvalue), ls="dashed", lw=2)