- RNeXML merge travis integration with support for rrdf. See pull #57
- Reply to Perry re nonparametric approaches
- 10am Andrew Skype meeting re: his manuscript
- rOpenSci discussion re: other services, formats.
- Reading Caldwell et al, Geophysical Letters. Good example of naiive data mining leading to spurious correlations from non-independent models. (/ht @FreshwaterEcology
- Looking over NSF’s Data Infrastructure Building Blocks, DIBB call and previously funded work.
- Remote control for Ubuntu via android? Considering:
- xbmc with xbmc-remote: Looks promising if a bit heavy; connection unsuccessful.
- WiFi Mouse successful connection via autoconnect and thethered wireless, works pretty smothly with limited setup. Does need root running
mouseserver, installed from an open source.debfile. - 2pm rOpenSci conference call
- Calculations, see below
Bayesian Early Warning
Continuing calculation from 2013/12/10, which integrates out \(\theta\). Next steps:
- Integrate out \(\sigma\)
- Grid-search \(\alpha\)
- Implement numerically
- Implement approach for time-heterogeneous case.
Simplifying and combing terms from where we left off: the integral of some Gaussian of variance \(\nu\) evaluates to \(\sqrt{2\pi \nu}\) which we solved for previously, giving us:
\[f(A_t) \left( 2 \pi V_t \right)^{-(T-1)/2} \left( 2 \pi V_t \right)^{1/2} \left(B^2(T-1) \right)^{-1/2}\]
\[ = f(A_t) \left( 2 \pi V_t \right)^{-(T-2)/2} \left(B^2(T-1) \right)^{-1/2}\]
Substituting in for \(V\) to have explicitly in terms of sigma, (note: Correcting the \(f(A_t)\) term, which was missing the denominator in the earlier version of 2013/12/10, now fixed).
\[ = \exp\left(\frac{}{}\right) \sigma^{-(T-2)} \left( \frac{\pi}{\alpha}\left(1- e^{-\alpha t}\right) \right)^{-(T-2)/2} \left(B^2(T-1) \right)^{-1/2}\]
and is then is integrable in terms of \(\sigma\) by recognizing this as a Gamma integral,
\[ = f(A_t) \sigma^{-(T-2)} \left( \frac{\pi}{\alpha}\left(1- e^{-\alpha t}\right) \right)^{-(T-2)/2} \left(B^2(T-1) \right)^{-1/2}\]