*** empty log message ***

diff --git a/src/ChangeLog b/src/ChangeLog
index 94709bd38d0ead7173df4cce01ea4de71af536c5..523a442f55bff1e839dd7d4caa891f4ce8b15b4d 100644 (file)
--- a/src/ChangeLog
+++ b/src/ChangeLog
@@ -1,42 +1,70 @@
-2003-06-17    <brouard@BROUARD>\r
-\r
-       * timeval.h (Module): Added included file to make use of\r
-       gettimeofday working on win32 with cygwin.\r
-\r
-2003-05-16  Brouard Nicolas  <brouard@brouard>\r
-\r
-       * imach.c (Module): Insert of a warning if the delay between two\r
-       waves is negative.\r
-\r
-       * imach.c (Module): \r
-\r
-2003-05-03    <brouard@BROUARD>\r
-\r
-       * imach.c: Some cleaning\r
-2003-03-28    <brouard@BROUARD>\r
-\r
-       * imach.c (Module): In version up to 0.92 likelihood was computed\r
-       as if date of death was unknown. Death was treated as any other\r
-       health state: the date of the interview describes the actual state\r
-       and not the date of a change in health state. The former idea was\r
-       to consider that at each interview the state was recorded\r
-       (healthy, disable or death) and IMaCh was corrected; but when we\r
-       introduced the exact date of death then we should have modified\r
-       the contribution of an exact death to the likelihood. This new\r
-       contribution is smaller and very dependent of the step unit\r
-       stepm. It is no more the probability to die between last interview\r
-       and month of death but the probability to survive from last\r
-       interview up to one month before death multiplied by the\r
-       probability to die within a month. Thanks to Chris\r
-       Jackson for correcting this bug.  Former versions increased\r
-       mortality artificially. The bad side is that we add another loop\r
-       which slows down the processing. The difference can be up to 10%\r
-       lower mortality.\r
-\r
-2002-05-30    <brouard@BROUARD>\r
-\r
-       * \r
-\r
-       * imach.c (Module): Add correlation matrix of one-step\r
-       probabilities and covariance matrix\r
-\r
+2003-06-17  Brouard Nicolas  <brouard@homebrou>\r\r
+\r\r
+       * imach.c (Repository): Check when date of death was earlier that\r\r
+       current date of interview. It may happen when the death was just\r\r
+       prior to the death. In this case, dh was negative and likelihood\r\r
+       was wrong (infinity). We still send an "Error" but patch by\r\r
+       assuming that the date of death was just one stepm after the\r\r
+       interview.\r\r
+       (Repository): Because some people have very long ID (first column)\r\r
+       we changed int to long in num[] and we added a new lvector for\r\r
+       memory allocation. But we also truncated to 8 characters (left\r\r
+       truncation)\r\r
+       (Repository): No more line truncation errors.\r\r
+\r\r
+2003-06-14  Brouard Nicolas  <brouard@homebrou>\r\r
+\r\r
+       * imach.c (Repository): Add of a routine likelione (likelihood\r\r
+       only once) which prints on a text file\r\r
+       (ilk) the contributions to the likelihood for each\r\r
+       individual/wave.\r\r
+\r\r
+2003-06-13  Brouard Nicolas  <brouard@homebrou>\r\r
+\r\r
+       * imach.c (Repository): Replace "freqsummary" at a correct\r\r
+       place. It differs from routine "prevalence" which may be called\r\r
+       many times. Probs is memory consuming and must be used with\r\r
+       parcimony.\r\r
+\r\r
+2003-06-17    <brouard@BROUARD>\r\r
+\r\r
+       * timeval.h (Module): Added included file to make use of\r\r
+       gettimeofday working on win32 with cygwin.\r\r
+\r\r
+2003-05-16  Brouard Nicolas  <brouard@brouard>\r\r
+\r\r
+       * imach.c (Module): Insert of a warning if the delay between two\r\r
+       waves is negative.\r\r
+\r\r
+       * imach.c (Module): \r\r
+\r\r
+2003-05-03    <brouard@BROUARD>\r\r
+\r\r
+       * imach.c: Some cleaning\r\r
+2003-03-28    <brouard@BROUARD>\r\r
+\r\r
+       * imach.c (Module): In version up to 0.92 likelihood was computed\r\r
+       as if date of death was unknown. Death was treated as any other\r\r
+       health state: the date of the interview describes the actual state\r\r
+       and not the date of a change in health state. The former idea was\r\r
+       to consider that at each interview the state was recorded\r\r
+       (healthy, disable or death) and IMaCh was corrected; but when we\r\r
+       introduced the exact date of death then we should have modified\r\r
+       the contribution of an exact death to the likelihood. This new\r\r
+       contribution is smaller and very dependent of the step unit\r\r
+       stepm. It is no more the probability to die between last interview\r\r
+       and month of death but the probability to survive from last\r\r
+       interview up to one month before death multiplied by the\r\r
+       probability to die within a month. Thanks to Chris\r\r
+       Jackson for correcting this bug.  Former versions increased\r\r
+       mortality artificially. The bad side is that we add another loop\r\r
+       which slows down the processing. The difference can be up to 10%\r\r
+       lower mortality.\r\r
+\r\r
+2002-05-30    <brouard@BROUARD>\r\r
+\r\r
+       * \r\r
+\r\r
+       * imach.c (Module): Add correlation matrix of one-step\r\r
+       probabilities and covariance matrix\r\r
+\r\r