Astsa

- added `SV.mle` to fit an SV model [with feedback (aka leverage) if desired] via quasi-MLE. Details are in the help file. `SVfilter` is now part of `SV.mle` and the original script is now in the xBox as `xSVfilter`.

- `tsplot` becomes more kick-ass with full control of the `Grid`

- `sarima` - prettified the output

- added `MEI` (Multivariate ENSO Index) data set

- `SV.mcmc` moved ESS display to the ACFs

- `test.linear` gets a title (`main`) control

- adjustment to `Kfilter(... , version=2)` - if the errors are _Q^½ w_t_ and _R^½ v_t_ then _S = cov(w_t, v_t)_, the _Q^½_ and _R^½_ are already included in the script. Often in this case _w_t = v_t_ and both are iid N(0, I) sequences, so `S = diag(1, q)`, the _q x q_ identity matrix.


- made `acf1` more flexible

- minor visual improvements to `SigExtract`

- increased the default max order and made detrend default on `spec.ic`

- improvements to `mvspec` like allow detrending via lowess and some other visual improvements like a gris-gris option, enhancing the voodoo grammar of astsa

- updated `cardox` to 2023

- added `ar.boot` to get the bootstrapped distributions of the parameters of a specified (by the order) AR model.

- added option in `trend` to plot (or not) the CIs - default is to plot (`ci=TRUE`).

- updated global temperature data sets and put all the old ones into the x box to be deleted (eventually) ... `gtemp_land`, `gtemp_ocean`, and `gtemp_both` are the updated/new sets

- updated `gnp` and `gdp` in sets `GNP` and `GDP`.

- added US population data (`USpop`) ... we thought it was in already, now it is

> **Note** There are a number of new scripts and some old ones are set to be retired eventually.

- Added [`autoSpec`](https://dx.doi.org/10.4310/21-SII703) for changepoint detection using local nonparametric spectra.

- Also added the data set `ENSO` which is the most recent update to SOI (an older version was used the paper that introduced autoSpec).

- The script uses the Bartlett kernel so it was made available for general use; see `bart`.

- Also added [`autoParm`](https://doi.org/10.1198/016214505000000745) for changepoint detection using local AR models.

✄   Some examples are in [FUN WITH ASTSA - structural breaks](https://github.com/nickpoison/astsa/blob/master/fun_with_astsa/fun_with_astsa.md#6-detecting-structural-breaks).

&128312;&128312;&128312;&128312;&128312;

- Added `Kfilter` and `Ksmooth` which are faster than the older `Kfilter0-1-2` and `Ksmooth0-1-2`, are easier to work with, and removes the need for 3 different scripts.



- Added `EM` which supersedes `EM0` and `EM1` and uses the quicker `Kfilter` and `Ksmooth` scripts. **In addition, the script now accepts inputs in both the state and observations equations.**


> __Warning__ the old script names `EM0-1`, `Kfilter0-1-2`, and `Ksmooth0-1-2` have an `x` in front of them now: `xEM0-1`, `xKfilter0-1-2`, and `xKsmooth0-1-2`. The scripts haven't changed (old scripts will still work with the `x` name change), but they will be phased out eventually. Converting code that used the old scripts to use the newer scripts should be easy with only a few minor changes in the call.

&128312;&128312;&128312;&128312;&128312;

- Updated (to run with the new `Kfilter` and `Ksmooth` scripts):

+ the Forward Filtering Backward Sampling script (`ffbs`) and
+ the simple univariate state space model (`ssm`) script

- Updated data files `gtemp_land` and `gtemp_ocean` to 2021

- Made `lag1.plot` and `lag2.plot` look more purty.

---
---

- Added `Months` to use with `pch` for monthly data; see the help file `?Months`.

- Tweaked `tsplot` by adding the ability to adjust the `mpg` graphics parameters settings (`?par`); see the help file `?tsplot`.

- A `tsplot` plot can now be stored by putting it in an object; e.g., `pl = tsplot(soi)`. Later, entering `pl` will restore the graph and it's possible to add to it (made possible by `recordPlot`).

- Added some Bayesian scripts (examples in [FUN WITH ASTSA](https://github.com/nickpoison/astsa/blob/master/fun_with_astsa/fun_with_astsa.md) - see the new section 9)
- Added `ar.mcmc` to fit AR models via Gibbs sampling
- Added `SV.mcmc` to fit stochastic volatility models
- ... and some financial data sets `sp500.gr` (S&P 500 daily returns) and `BCJ` (returns of 3 banks)
- Added `ffbs` (forward filter backward sample algorithm) for linear state space models
- Added `ESS` to estimate the effective sample size

- Added a line to `detrend` to make sure the input series is univariate (already there in `trend`). Also, in `trend`, forgot to add the span option for lowess (actually `stats::loess` with a robust option) - this has been corrected.

- Added the ability to change the legend text color in `lag1.plot` and `lag2.plot` and set the default to black - it makes the values easier to see, especially if the background of the legend is transparent.


---

+ Added two new scripts detrend() and trend(). The first one returns a detrended series using a polynomial regression (default is linear) or lowess (with the default span). The second script fits a trend (same options as detrend) and produces a graphic of the series with the trend and error bounds superimposed. The trend and error bounds are returned invisibly.

---
Versions 1.14 - Sept 2021

+ Added sleep state and movement data (`sleep1` and `sleep2`) - more details in the help files.

+ Added option to specify a kernel in `specenv` and if `spans` and `kernel` are both `NULL`, the spectral envelope will be based on the periodogram. Also changed the way it checks if `section` is a proper sequence and added option to taper the data prior to estimating spectra.

+ Some minor changes:

- In `matrixpwr` changed `isSymmetric(A)` to `isSymmetric(unname(A))` because a symmetric matrix is not taken as such if the column and row names are not the same.

- In `arma.spec` if there is near parameter redundancy, `ylim` is now adjusted so the figure will be close to the white noise (uniform) density.

+ v1.14 (Sept 2021) Just in time for a new skool year - v1.14 is on CRAN - it is v1.13.2 with minor changes to please the CRAN gods.

---
Versions 1.13 - May 2021

+ Also, changed the LAG axis labels on `acf1`, `acf2`, and `ccf2` to show the
frequency of the series if it's bigger than one. For example, `soi` has
frequency 12 and the LAG axis of `acf1(soi)` will be ticked as 1, 2, 3, ...
but the label now emphasizes that each tick is LAG divided by 12.

+ v1.13.1 (July 2021) Some minor improvements to `tsplot-spaghetti`, `sarima.sim`, `sarima`, and `arma.spec`.

+ For `sarima.sim`, I forgot to add the `innov` argument in the call (only a problem if you wanted to use your own innovations), but it seemed to work in the examples I tried (maybe it got passed in ...) ¿Quién sabe? Or maybe it was just voodoo.

+ Updated some man page (adding sources and subtracting typos).

+ v1.13 is on CRAN. There are lots of additions to the package that are listed below.


---

Versions 1.12 - started Dec 2020