[BACK]Return to ChangeLog CVS log [TXT] [DIR] Up to [Local Repository] / imach / src

Diff for /imach/src/ChangeLog between versions 1.3 and 1.6

version 1.3, 2003/05/16 10:46:40 version 1.6, 2003/06/17 20:16:05
Line 1 Line 1
 2003-05-16  Brouard Nicolas  <brouard@brouard>  2003-06-17  Brouard Nicolas  <brouard@homebrou>
   
         * imach.c (Module):          * imach.c (Repository): Check when date of death was earlier that
           current date of interview. It may happen when the death was just
 2003-05-03    <brouard@BROUARD>          prior to the death. In this case, dh was negative and likelihood
           was wrong (infinity). We still send an "Error" but patch by
         * imach.c: Some cleaning          assuming that the date of death was just one stepm after the
 2003-03-28    <brouard@BROUARD>          interview.
           (Repository): Because some people have very long ID (first column)
         * imach.c (Module): In version up to 0.92 likelihood was computed          we changed int to long in num[] and we added a new lvector for
         as if date of death was unknown. Death was treated as any other          memory allocation. But we also truncated to 8 characters (left
         health state: the date of the interview describes the actual state          truncation)
         and not the date of a change in health state. The former idea was          (Repository): No more line truncation errors.
         to consider that at each interview the state was recorded  
         (healthy, disable or death) and IMaCh was corrected; but when we  2003-06-14  Brouard Nicolas  <brouard@homebrou>
         introduced the exact date of death then we should have modified  
         the contribution of an exact death to the likelihood. This new          * imach.c (Repository): Add of a routine likelione (likelihood
         contribution is smaller and very dependent of the step unit          only once) which prints on a text file
         stepm. It is no more the probability to die between last interview          (ilk) the contributions to the likelihood for each
         and month of death but the probability to survive from last          individual/wave.
         interview up to one month before death multiplied by the  
         probability to die within a month. Thanks to Chris  2003-06-13  Brouard Nicolas  <brouard@homebrou>
         Jackson for correcting this bug.  Former versions increased  
         mortality artificially. The bad side is that we add another loop          * imach.c (Repository): Replace "freqsummary" at a correct
         which slows down the processing. The difference can be up to 10%          place. It differs from routine "prevalence" which may be called
         lower mortality.          many times. Probs is memory consuming and must be used with
           parcimony.
 2002-05-30    <brouard@BROUARD>  
   2003-06-17    <brouard@BROUARD>
         *  
           * timeval.h (Module): Added included file to make use of
         * imach.c (Module): Add correlation matrix of one-step          gettimeofday working on win32 with cygwin.
         probabilities and covariance matrix  
   2003-05-16  Brouard Nicolas  <brouard@brouard>
   
           * imach.c (Module): Insert of a warning if the delay between two
           waves is negative.
   
           * imach.c (Module):
   
   2003-05-03    <brouard@BROUARD>
   
           * imach.c: Some cleaning
   2003-03-28    <brouard@BROUARD>
   
           * imach.c (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.
   
   2002-05-30    <brouard@BROUARD>
   
           *
   
           * imach.c (Module): Add correlation matrix of one-step
           probabilities and covariance matrix
   
Removed from v.1.3  
changed lines
  Added in v.1.6

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>