GammaLib

Performing further cross checks between the fermi science tools and ctools the following problem came up:

If one takes the xmlfile from the science tools and tries to analyse it with gammalib/ctools the following happens:

input example:

<pre>
<spectrum normPar="norm" type="LogParabola">
<spectrum normPar="norm" type="LogParabola">
<parameter error="0.0403546564" error=&quot;0.5109914278&quot; free="1" max="100000" max=&quot;1000000&quot; min="1e-05" name="norm" scale="6.664498894e-12" value="0.330797703"/> scale=&quot;1e-12&quot; value=&quot;1.427987392&quot;/&gt;
<parameter error="0.1932913856" error=&quot;0.6780729296&quot; free="1" max="5" min="-0" min=&quot;0&quot; name="alpha" scale="1" value="2.762437733"/> value=&quot;3.418006945&quot;/&gt;
<parameter error="0.9494770106" error=&quot;0.2345999029&quot; free="1" max="10" min="-10" max=&quot;1&quot; min=&quot;-1&quot; name="beta" scale="0.1" value="3.304939119"/> scale=&quot;1&quot; value=&quot;0.6208371083&quot;/&gt;
<parameter free="0" max="300" min="0.02" free=&quot;1&quot; max=&quot;1000&quot; min=&quot;0.01&quot; name="Eb" scale="1000" value="1.444281016"/> value=&quot;1.6254829&quot;/&gt;
</spectrum>
</pre>

this means if the source is fitted also the Pivotenergy is free, which leads to the huge errors on the fitted parameters ( I think due to problem of correlated parameters) an example shown here:
<pre>
<spectrum type="LogParabola">
<parameter name="Prefactor" value="0.001" error="0.0510866" scale="1e-09" min="0.001" max="1000" free="1" />
<parameter name="Index" value="-3.71444e-05" error="878492" scale="1" min="-10" max="10" free="1" />
<parameter name="Curvature" value="-0.319421" error="0.108217" scale="1" min="-10" max="10" free="1" />
<parameter name="Scale" value="511.743" error="7.03715e+08" scale="1" min="20" max="10000" free="1" />
</spectrum>
</pre>

But also fixing the Pivotenergy does not solve the problem at the moment since the definitions of the parameters in the fermi science tools and gammalib/ctools are not completely compatible. This leads to a fit result like this:

<pre>
<spectrum type="LogParabola">
<parameter name="Prefactor" value="1.13197" error="0.0615543" scale="6.6645e-12" min="1e-05" max="100000" free="1" />
<parameter name="Index" value="-4.18061" error="0.0833878" scale="1" min="-5" max="5" free="1" />
<parameter name="Curvature" value="-10" error="0" scale="0.1" min="-10" max="10" free="1" />
<parameter name="Scale" value="1.44428" scale="1000" min="0.02" max="300" free="0" />
</spectrum>

</pre>