CVS log for imach096d/src/imach.c

Up to [Local Repository] / imach096d / src

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

Version 0.96d. Population forecasting command line is (temporarily)
suppressed.

* imach.c (Repository): Errors in subdirf, 2, 3 while printing tmpout is
rewritten within the same printf. Workaround: many printfs.

* imach.c (Repository):
(Repository): Using imachwizard code to output a more meaningful covariance
matrix (cov(a12,c31) instead of numbers.

Just cleaning

(Module): On windows (cygwin) function asctime_r doesn't
exist so I changed back to asctime which exists.
(Module): Version 0.96b

(Module): On windows (cygwin) function asctime_r doesn't
exist so I changed back to asctime which exists.

* imach.c (Repository): Duplicated warning errors corrected.
(Repository): Elapsed time after each iteration is now output. It
helps to forecast when convergence will be reached. Elapsed time
is stamped in powell.  We created a new html file for the graphs
concerning matrix of covariance. It has extension -cov.htm.

(Module): Some bugs corrected for windows. Also, when
mle=-1 a template is output in file "or"mypar.txt with the design
of the covariance matrix to be input.

(Module): Some bugs corrected for windows. Also, when
mle=-1 a template is output in file "or"mypar.txt with the design
of the covariance matrix to be input.

* imach.c (Repository): Create a sub-directory where all the secondary files are. Only imach, htm, gp and r(imach) are on the main directory. Correct time and other things.

Version 0.96

(Module): Change position of html and gnuplot routines and added
routine fileappend.

* imach.c (Repository): Check when date of death was earlier that
current date of interview. It may happen when the death was just
prior to the death. In this case, dh was negative and likelihood
was wrong (infinity). We still send an "Error" but patch by
assuming that the date of death was just one stepm after the
interview.
(Repository): Because some people have very long ID (first column)
we changed int to long in num[] and we added a new lvector for
memory allocation. But we also truncated to 8 characters (left
truncation)
(Repository): No more line truncation errors.

* imach.c (Repository): Replace "freqsummary" at a correct
place. It differs from routine "prevalence" which may be called
many times. Probs is memory consuming and must be used with
parcimony.
Version 0.95a2 (should output exactly the same maximization than 0.8a2)

We decided to release, in June 2003, a version 0.8a1 (May 2002) of
former version 0.8a to correct a bug highlighted by Chris Jackson.
It can be useful until version 0.9x will be released.

In version 0.8a, likelihood was computed as if date of death was unknown. Death
was treated as any other health state, i.e the date of the interview
describes the actual state and not the date of a change in health
state.

It means that in the early version of IMaCh we were considering that
at each interview the current state was recorded (healthy, disable or
death).  And IMaCh was correct. But, when we introduced the exact date
of death then we should have modified the contribution of an exact
(more precise) date of death to the likelihood. This new contribution
is smaller and very dependent of the step unit stepm. It is no more
the probability to die between last interview and month of death
(which is the way version 0.8a of IMaCh was doing) but the probability
to survive from last interview up to one month before death multiplied
by the probability to die within a month. Thanks to Chris Jackson for
correcting this bug.  Former versions increased mortality
artificially. The bad side is that we add another loop which slows
down the processing. The difference can be up to 10% lower mortality.

Minor changes to run version 0.8a on Linux. Bad ifdef linux.

*** empty log message ***

Add log in  imach.c and  fullversion number is now printed.

*** empty log message ***

Trying to add the true revision is the program and log

*** empty log message ***

minor changes

(Module): Insert of a warning if the delay between two waves is
negative.

(Module):

Some cleaning

Check memory allocation through valgrind on Linux and libnjamd .
So it is free of malloc bugs now!
Version 0.95

In freqsummary, the observed prevalence is now computed from all the
waves and the bug in the forecasting routine is corrected. We can
forecast prevalence if ncov=0.

(Module): In version up to 0.92 likelihood was computed
as if date of death was unknown. Death was treated as any other
health state: the date of the interview describes the actual state
and not the date of a change in health state. The former idea was
to consider that at each interview the state was recorded
(healthy, disable or death) and IMaCh was corrected; but when we
introduced the exact date of death then we should have modified
the contribution of an exact death to the likelihood. This new
contribution is smaller and very dependent of the step unit
stepm. It is no more the probability to die between last interview
and month of death but the probability to survive from last
interview up to one month before death multiplied by the
probability to die within a month. Thanks to Chris
Jackson for correcting this bug.  Former versions increased
mortality artificially. The bad side is that we add another loop
which slows down the processing. The difference can be up to 10%
lower mortality. Version 0.93

(Module): In version up to 0.92 likelihood was computed
as if date of death was unknown. Death was treated as any other
health state: the date of the interview describes the actual state
and not the date of a change in health state. The former idea was
to consider that at each interview the state was recorded
(healthy, disable or death) and IMaCh was corrected; but when we
introduced the exact date of death then we should have modified
the contribution of an exact death to the likelihood. This new
contribution is smaller and very dependent of the step unit
stepm. It is no more the probability to die between last interview
and month of death but the probability to survive from last
interview up to one month before death multiplied by the
probability to die within a month. Thanks to Chris
Jackson for correcting this bug.  Former versions increased
mortality artificially. The bad side is that we add another loop
which slows down the processing. The difference can be up to 10%
lower mortality.

Preforecast routine cleaned. Change of notation, comments added.
Version 0.92

*** empty log message ***

Correction of initialisations in case of several death states.

Fixed probability of death over estepm months. We don't calculate the
probability of death over a year anymore. Fix estepm as 12 and you
will get it.

I changed some comments

Correction of bug related to the covariates

Extrapolation doesn't seem to work fine, so we decided to use
interpolation only ie at the cost of an additional matrix product if
the bias, bbh, is positive, in order to make it negative
(interpolation).
We keep other methods mle=4 no adjustment for bias, mle=3 inter-extra
exponential, mle=2 inter-extra linear.

Trying a linear interpolation extrapolation in the computing of
likelihood for large stepm.

just a sign error in mlikeli

mle=1 corrects bias
mle=2 no correction

Allocation bug

Major change: likelihood is modified in order to unbiased the duration
between 2 waves which is rounded (biased) to the nearest multiple of
stepm.
Some cleaning of the code. Many allocations were not freed correctly.

Comments added and corrections in routine moving average done.

*** empty log message ***

*** empty log message ***

Bug in ***movingaverage

mobilav can be 3, 5 or 7 instead of only 1 (which defaults to 5).

Moving average of observed prevalence moved

Some improvments

Some errors corrected in diagonalisation of covariance-matrice

*** empty log message ***

Changes in the htm file.

version 0.8h

Many changes:
(a) We look at covariance between step probabilities by drawing
confidence interval ellipsoids of two step probabilities pij versus
pkl.
It is done by diagonalizing the inverse of the convariance matrix.

(b) In order to write to the gnuplot file or even to the html file
without using to much memory we are now reopening ficgp and fichtm
with an "append" option each time we need them into the main program.
It is better than keeping matrices in memory until the end.

(Module): Add correlation matrix of one-step
probabilities and covariance matrix

The bug in printinghtm is corrected

Change version number ot 0.8e

Change of gif to png.
Clarify some of the outputs.
Add a new graph: pij*YEARM/stepm in order to get pseudo-incidence per
year.

Change gif images to png format because of gif copyright.
Change wgnuplot to gnuplot
Correct a small bug

Version 0.8a. Standard deviation of health expectancies eij.

*** empty log message ***

*** empty log message ***

The bug in printinghtml is corrected.

Changed  version to 0.8b

Estepm added. We are calculating Life expectancy by interpolating the
survival  function pijx linearly every estepm month. Estepm should be
higher than stepm. It is a new parameter (sorry).

Version used for article20.tex Changes in varevsij.

CHANGE ncov to ncovcol
There was a confusion with older ncov parameter. In fact it was the
number of columns, between id and date of birth, which can be used for
covariates. In the program we use ncovmodel for the real number of
covariates. Version 0.8 !

Some cleaning in esij and nhstepm

Supress a bug in esij

0.71a cleaning done

Version 0.71

Output mle=0 in the Rparameter file so thart clicking on
rparameterfil.imach will skip the optimization process.

Some improvements in forecasting and the bug relating to agemax is
corrected.

Some changes in population forecasting

Add date in the freqsummary file.

Cleaning of the main function: gnuplot and html file are now
subroutines; the bug in the forecasting is corrected; the moving
average statement moved to the last command line.

*** empty log message ***

Merge OK

First merge between Agn�ès and Nicolas!!

Error 108 if stepm is not equal to 1 (month) because forecasting
is only possible with a former optimization with a step of one month.

Prevalence and forecasting

Forecasting and calculation of the mean date of interview

Forecasting

Some bugs corrected

Version 0.7
hPijx are now computed for h=1 year instead of 2 years

Moving average  on cross-sectional prevalences

Forecasting

Variance of one-step transitions added and forecasting of prevalences

Remise �à jour du serveur CVS peu de changements

Updated to Windows2000: changes in the path name to execute wgnuplot.

Changes in calculation spaces between two waves

Change of index.htm to parameterfile.htm
Version 0.64b

Changes index.htm: title, model, ...

Add age*V1

some changes index.htm
Version16.

suppressing cygwinpath. split directory instead.
version 15

*** empty log message ***

version 13

Version Rostock: covariables + graphique proba elementaires

Import de imach064

Initial revision