Note: variables in tvc equation interacted with ln(_t)

[eric some]

Dear forumists,
I have run a Cox regression analysis using the "stcox" command and "tvc" and "texp" options. I got this note: "variables in tvc equation interacted with ln(_t)". What do these interactions imply for the analysis? Do I need to introduce some interactions in my model? What kind of action is needed when you have this kind of note?
Another issue I faced is that I got in some models disproportionately small (ex: 5.05e-10 (95% CI:4.36e-12; 5.84e-08)) or big (174362.3 (95% CI:35142.78; 865105.4) hazard ratios.
Does this mean that there is something wrong with my data or the analysis? If not, how do I interpret this? Am I allowed to replace 5.05e-10 by 0? What about HR=174362.3?
Thanks in advance for your help!

[Steve Samuels]

1. stcox created the interactions for you; you don't need to do anything.

2. Please show us your stcox command and all the output from a model with near-zero estimates. (This is requested in the FAQ). Describe the variable(s) with the small HRs.

[eric some]

Dear Steven,
thanks so much for your response. I have attached a file with pictures showing the stcox command and some output from two country-stratified models.
The variables causing problems are:
- "bfinitime" for breastfeeding initiation time with the reference group initiating breastfeeding within one hour postpartum, the second group after one hour but within one day postpartum and the 3rd group after one day postpartum.
- "deliv" for type of delivery with the reference group delivering by vaginal route and the second group delivering by c-section.

the "bfinitime" variable shows very small HR in the first country (country-stratified models) and a huge HR in the second country.
the "deliv" variable shows also a very small HR in the second country.
What would these outputs mean for my data and how should I interpret them? I have never seen such extreme results in a paper!

[Steve Samuels]

Please put commands and output between CODE delimiters as FAQ Section 12 asks. Photos are nearly unreadable and one can't copy and paste from them.

[eric some]

Here is the command:
[stcox i.tertile i.rand i.agegroup i.educ i.marital i.occup i.parity i.deliv i.bfinitime, ///
tvc(rand agegroup educ marital occup parity deliv bfinitime) texp( ln(_t) ) ]
and here part of the output:
[ _t | Haz. Ratio Std. Err. z P>|z| [95% Conf. Interval]]
[ 2.deliv | 5.05e-10 1.22e-09 -8.83 0.000 4.36e-12 5.84e-08|
bfinitime |
2 | .9061764 .028648 -3.12 0.002 .8517316 .9641014
99 | 174362.3 142491.8 14.77 0.000 35142.78 865105.4]

The variables causing problems are:
- "bfinitime" for breastfeeding initiation time with the reference group initiating breastfeeding within one hour postpartum, the second group after one hour but within one day postpartum and the 3rd group after one day postpartum.
- "deliv" for type of delivery with the reference group delivering by vaginal route and the second group delivering by c-section.

the "bfinitime" variable shows very small HR in the first country (country-stratified models) and a huge HR in the second country (showed in the code above).
the "deliv" variable shows also a very small HR in the second country (see code above).
What would these outputs mean for my data and how should I interpret them?

[eric some]

[ _t | Haz. Ratio Std. Err. z P>|z| [95% Conf. Interval]]
[ 2.deliv | 5.05e-10 1.22e-09 -8.83 0.000 4.36e-12 5.84e-08|
bfinitime |
2 | .9061764 .028648 -3.12 0.002 .8517316 .9641014
99 | 174362.3 142491.8 14.77 0.000 35142.78 865105.4]

[Stephen Jenkins]

Eric: the Stata material in your last two posts is not properly readable. Please follow Steve Samuels' excellent advice in #4

[eric some]

Yes I know. But I have read the FAQ 12 and tried to apply the instructions. Despite careful manual arrangement of the data, they appear disordered like this once I click to post them. I shall try once more.
Thanks

[eric some]

Sorry. The preview is not that better.

[Steve Samuels]

Put the text between CODE delimiters, i.e. "[C O D E]" and "[/C O D E]" (but remove the spaces-the forum software doesn't allow us to actually quote the delimiters. . Alternatively, hit the "A" button ("Advanced Editor") in the upper right of the edit window. Select your text, then hit the "#" button.