Jan,

Suppose that you have a subject who has the following event and covariate structure:

Time Status Covariate Observed 0 OK 6 Yes 1 OK 3 Yes 2 OK 4 Yes 3 Failed . Yes

You know that at T=3 the subject failed, but you cannot include this observation in your analysis because the covariate value was missing. The only information you can use in your time- dependent covariates regression are the first three observations on this subject. Through the third observation, you know that the subject had not failed. You cannot employ the fourth observation because of the missing covariate value. Hence, you effectively have the following data for this subject:

Time Status Covariate Observed 0 OK 6 Yes 1 OK 3 Yes 2 OK 4 Yes

You would immediately conclude that this subject was censored. As of the last usable information, the subject had not failed. If you did not have subsequent missing data when the subject failed, then you would have recorded the failure as an event. But because of the missing value for the time-dependent covariate and the inability to use the observation because of the missing value, you are not able to record the failure. This is what the note in your log file is indicating.

Just out of curiousity, do you have recorded failures where the time-dependent covariate is non-missing? If the time-dependent covariate is missing for all your failures because you can only collect the covariate when the subject is vital, then you probably need to lag your time-dependent covariates so that you use the information collected at the previous vital measurement to predict status at the next occassion. That is, you should reconstruct your data so that you have

Time Status Covariate Observed 1 OK 6 Yes 2 OK 3 Yes 3 Failed 4 Yes

HTH,

Dale

--- Jan Feyzi <jmfeyzi@FACSTAFF.WISC.EDU> wrote:

> I am including some time-dependent covariates in my model. Some of > them > have missing values. In the .log file, there is the following > statement: > > NOTE: 35 observations that were events in the input data set are > counted > as censored observations due to missing or invalid values > in the time-dependent explanatory variables. > > This is very surprising to me - in the model without the > time-dependent > terms, records with missing covariates are not included in the > analysis. > > Anyone know why these values are counted as censored? > > Thanks in advance, > > -- Jan >

===== --------------------------------------- Dale McLerran Fred Hutchinson Cancer Research Center mailto: dmclerra@NO_SPAMfhcrc.org Ph: (206) 667-2926 Fax: (206) 667-5977 ---------------------------------------

__________________________________ Do you Yahoo!? Jazz up your holiday email with celebrity designs. Learn more. http://celebrity.mail.yahoo.com