Some restrictions for the use of the function sympostest() lifted.
Rapp now compiled with Microsoft Visual C++ 2017. I cannot
see an increase in speed of execution from the 2015 version,
but there might be other benefits.
Function sympostest() introduced in Proc Mbasic.
Proc Taran is now somewhat faster for a large number (greater
than about 1250) of free parameters, due to meticulous choice of
inversion algorithm depending on the size of the Fisher information
One more matrix inversion algorithm added in Proc Matrix and
Proc Mbasic - the Householder method, which might be useful
for some dense matrices.
Recently I waded out in too deep water for me by trying to
parallelize matrix multiplication. I just discovered that it
had unwanted consequences, so all forced parallelization is now
removed. If you downloaded the Rapp exes in September you
might have obtained unwanted results.
Proc Mbasic has now more printing options.
In Proc Mbasic a bug concerning output of arrays to print files
has been corrected.
Some complex operations made faster and/or more accurate after
implementation of "Numerical Recipes in C", § 5.4. Complex
square root added.
In Proc Mbasic a bug, that disabled Redim and declarations with
variables as dimension limits for complex arrays, has been fixed.
Array and matrix operations in Proc Mbasic extended to complex
arrays. Only LU decomposition is presently implemented for
complex matrix inversion. The keyword complex is now required
for complex variables and arrays - eg "mouble A complex"
is now not the same as "mouble A". There were too many
ambiguities to admit the new functionality without the requirement.
Arithmetical operations for arrays beyond matrices introduced
in Proc Mbasic.
Some malfunctions (not exactly bugs) fixed in Proc Mbasic.
Some more bugs fixed in Proc Mbasic, concerning negations
of matrices and complex variables.
A bug fixed, which affected running Rapp with parameter CHA.
Type Rapp /? in the command prompt for parameters to Rapp.
Yet another bug fixed in Proc Mbasic, which made identifier
names with only one character sometimes invalid.
Proc Ddist has been much generalized to sums of independent
but not necessarily identically distributed random variables
with a discrete distribution. Multiprecision computation of
values and probabilities is also enabled.
I put the source code out in the open as a zip file and an
open letter on the maintenance of Rapp.
Proc Ddist has been somewhat generalized.
Proc Chaall now handles a practically unlimited number of
changes and text files with practically unlimited line length.
New Proc Ddist for the distribution of a sum of IID random
variables with a discrete distribution.
A bug fixed in Proc Mbasic / bindist() and bindistc().
Proc Calend, which has existed a few years, is now described
in the manual.
There is now a guest book at the bottom of the Rapp page.
I intend this not so much for communication between users
and myself, since anyone can write an e-mail to me and I
might not look at the guest book for several days. Rather I
intend it for users to share tips between them.
Appendix 2 in the manual is now named Cps4manual.Pdf.
An error in the expression for bias-corrected risk premium
on p. 24 now denoted (C5) has been corrected. The article
is reworked for more clarity and ease of understanding.
The cubic equation x3 + A x2 + B x + C = 0 can now be
solved with the function cube() in Proc Mbasic.
A bug fixed in Proc Mbasic / read statement when several
variables are read from the same input line.
A bug fixed in Proc Mbasic / sort statement. A mouble
array with both positive and negative elements was
incorrectly sorted before this correction.
Complex variables can now be treated in Proc Mbasic.
Please notify me if you find bugs.
A bug fixed in Proc Mbasic / gamdistinvc().
Some more variants of inverse functions added.
Inverses for the Γ- and χ2 distributions now in Proc Mbasic.
Inverses for the beta-, binomial and F-distributions available
in Proc Mbasic. The t inverse har now full mouble precision.
Proc Linreg has been augmented to give βj coefficients under
the assumption that α = 0.
A bug fixed in Reserves / Files and folders / Sort and complete
and in the template program TemplateR.rpp. Thanks to Rinjani!
In Proc Mbasic I have changed the order of the arguments
for fdist() and fdistc(). The t-distribution is now named
tdist() and its inverse tdistinv(). The inverse is now more
exact. New names for the β and binomial distributions.
The binomial and Poisson distribution functions added to
The χ2 and β distribution functions added to Proc Mbasic.
Modified Bessel functions Iv(x), Kv(x) added to Proc Mbasic.
More flexibility in Mbasic array handling.
A bug in Mbasic handling of conditions corrected. More
flexibility in conditional expressions.
The same Bessel functions now available in mouble with max
precision 306 decimal digits.
Bessel functions Jv(x) and Yv(x) now available in Proc Mbasic,
although at present only in double.
ψ(x) = Psi(x) = d log(Gamma(x))/dx = Gamma'(x)/Gamma(x)
now available in Proc Mbasic.
The Gamma distribution function now available in Proc Mbasic.
Slightly enhanced functionality in the print and read statements
of Proc Mbasic. More explanations of how the Gamma function
Rapp32Vc2015.Exe replaces Rapp32Vc2013.Exe due to new
compiler and linker.
A bug in Gam() and Loggam() in Proc Mbasic corrected.
(Since the proc is so new, it is unlikely that all bugs have been
corrected by now. Any help to discover more bugs would be
The Γ function is now available in mouble in Proc Mbasic,
however with maximal precision 306 decimal digits.
Another bug corrected in Proc Mbasic.
Bugs corrected in Reserves / RDC. Thanks to Rinjani Pebriawan!
Some improvements in Rappmenus / Reserves following
experiences by Rinjani Pebriawan at Universitas Gadjah Mada,
Imperfections in Mbasic handling of conditions corrected.
Variants of Rapp.Exe for up to 9000 correct decimal digits
now available in Rapp All-In-One zip file
New Proc Mbasic. Look in the manual. Excerpts:
Mbasic is a mathematical language within the language Rapp.
M in Mbasic stands for Multiple precison, ie you choose the
number of correct decimal digits up to 306 for a data type
called mouble. The common double data type with an 8-byte
storage is also available. Also M stands for Matrix Basic,
since it has powerful matrix operations. You can largely use
the same syntax as in classical Basic (such as For / Next,
Gosub etc). Some C syntax is also employed.
New choice in Rappmenus /Reserves / Files and folders to let
Rapp make a line with Paydt = Reportdt for each claim if none
exists. The files will be sorted on ID also.
More explanations on grouping in quantile intervals in RDC,
in Appendix 6 of Rappmane.doc and the sheet Paid in Excel.
2015-09-12. Yet another bug in RDC, discovered while working
with Mujiati and Ari, has been corrected. Thanks to them again!
MSEP now yet somewhat smaller.
2015-09-08. After the invention of a superfast method of sorting,
the execution time for RDC mentioned below is reduced so that
it is now about 26 % longer than before with many quantiles.
A bug in the computation of quantiles, affecting Proc Percen
and the computation of reserves with the RDC method, has
been corrected. Thanks to Mujiati Dwi Kartikasari and Ari
Dwi Hartanto at Universitas Gadjah Mada, Yogyakarta,
Indonesia, who discovered the bug! Execution time for RDC
is now somewhat longer by about 30 %, which is noticeable
in BICH. On the other hand, the more precise computation of
quantiles gives a smaller MSEP for RDC.
Still newer pseudo-estimators in credibility. See manual,
A bug corrected in credibility.
Somewhat new pseudo-estimators in credibility. The manual, its
Appendix 2 and the output from Rapp / Proc Taran are now
consistent with each other, I hope.
New parameters Leftfoot(), Rightfoot(), Lefthead(), Righthead()
for Procs Graf and Map. They display on all pages.
New parameter Landscape for Procs Graf and Map. Also as
checkboxes in the menus.
Parameter sort() in Proc Data now accepts /D for descending.
Yet another bug with parameter quick corrected. I hope I have
now corrected all bugs there, and most bugs elsewhere in Rapp.
Parameter quick or q to Procs Match and Sum is reinserted.
The bug mentioned below has been corrected.
Rapp /? on the command line now gives usages.
Some "features" (malfunctioning in extreme cases) have been
corrected in procs Map and Svg2co.
New submenu Svg2co in menu Maps.
The time measurement of procs is augmented so that both the
calendar time elapsed and consumed CPU are stated.
Rapp is now made with MSVC 2013. The name of the 32-bit Rapp
is changed from Rapp32Vc2010e.Exe to Rapp32Vc2013.Exe.
Info for those interested in the forthcoming Rapp Open Source
project: I have tried the Intel C® Compiler and its linker. They
did not give faster execution speed than Microsoft's compiler and
linker MSVC. Since all other compilers are said to be slower than
MSVC, this means that Rapp will continue to be made with MSVC.
TempMonthCl modified to delete only files older than 14 days.
This will eliminate the risk of collisions between two or more
Rapp programs run simultaneously, where one will create temp
files and another one will try to delete them before they have
Parameter TempMonthCl for monthly cleanup of tempmapp added
to Proc Init.
German language - at lan(d) in Proc Init - now also for BICH and
Reserves. Please notify me if you spot some error in my German.
Some improvement in the layout of the menus in Rappmenus.
From the experiences in the Tariff Analysis workshop November
2014 in Bandung and Yogyakarta, Indonesia:
Better documentation and elimination of the nuisance x as second
argument to Rapp in the Command Prompt. The x could be
eliminated through an environment variable i Br.Bat.
Parameter quick or q to Procs Match and Sum deleted. I discovered
it caused bugs for large files in memory.
More functionality in the menu Maps, with roads etc.
New Proc Ovelim for eliminating policy period overlaps.
New functionality for the procs Coofil and Svg2co that make
coordinate files for Proc Map.
Reserving manual Resvmane.doc has been updated.
A feature, concerning use of a Schnieper exposure file and premium
file for Benktander and Cape Cod at selection by segments, has been
augmented. Affects menus "BICH stochastic reserving", "Reserves"
and "Triangle based simple reserves".
A simple menu for reserves, using aggregated triangle data, is
Liu & Verrall (2009) prediction errors reinserted in procs Restri
and Bich. Aggregated triangle or list data can now be used in
menu Reserves. (Press button "Files and folders".) The BICH
prediction errors are the best if more than a few claims are
reported in the first development period, but I recognize that
BICH is not always possible.
Rapp.rar deleteted. Use rappexes.htm to download new exes.
A bug in the menus' graph-making corrected.
A menu Maps is now available in Rappmenus.Exe.
The Rapp package is now available in rar form on the AllInOne page.
Input from text files augmented so that all files, including such that
might have been transferred from Mac or Unix, are read corerctly.
New Proc Matrix for matrix determinant computation, inversion,
multiplication and transpose.
Non-pseudo tau²-estimators in my credibility method. Search
2014-02-26 in Rappmane.doc.