News for Package 'ACDm'

Changes in ACDm version 1.1.0 (2025-07-16)

• Added the following models for acdFit(): EXACD, AACD, BCACD, TACD, TAMACD.

• Added optimization with analytical derivatives in acdFit(). Works for all models, but currently only with dist = "exponential" and not with exogenous variables.

• Added support for broom::tidy and broom::glance.

• Fixed a bug that prevented exogenous variables from being used with any distribution other than the exponential (thanks to Mark Rotter for pointing this out).

• Fixed a minor bug in plotLL().

• Fixed a bug in diurnalAdj() where method = "FFF" with aggregation = "all" previously returned the deterministic part instead of the adjusted durations.

• Fixed a bug in predict.acdFit()—previously, predictions could only be made from a fitted ACD(1,1) model with exponential errors (thanks to Renova for pointing this out).

• Fixed a bug in rgengamma() (thanks to Steinar Veka for pointing this out).

Changes in ACDm version 1.0.4.2 (2022-11-14)

• Minor fix to make comments in C code valid UTF-8.

Changes in ACDm version 1.0.4 (2016-07-15)

• One can now use exogenous regressors when fitting ACD models using acdFit(). See the function's help page.

• Fixed a bug in the internal C functions getLL_ACDcall, getLL_LACD1call, and getLL_LACD2call that unabled estimating ACD(p, q), LACD1(p, q), and LACD2(p, q) models when p differed from q.

• Added the predict() method.

• Added the plotLL() function.

• Changed how the trace argument inside the control list in the acdFit() function works. It will now plot the MLE search path, recorded as the parameter values each time the optimization function calls the log likelihood.

Changes in ACDm version 1.0.3 (2015-12-01)

• The examples were set to "dontrun" for a few of the functions.

• Changed the format of the time columns in the example data durData and adjDurData from POSIXlt to character strings. The time zone feature of the POSIXlt format made some of the examples to fail in other time zones - preventing the package to be installed from source within those time zones.

For the function computeDurations():

• The function now works properly for sub-second precision.

• Fixed bug where the duration of the second transaction of a day would sometimes be calculated as the duration from the opening time

• Fixed bug where the last duration would sometimes not be added.

• Having only one of the price/volume column should not work.

• Now checks if any negative durations were computed - if so, checks that the data were given in chronological order.

For the function diurnalAdj():

• for method = "cubicSpline": now checks if the time stamps of the durations are within the range of the nodes argument.