why does my output come out as 2.89e+09 - is there a remedy for this?

[Tara Boyle]

Hi why is it that my output from my hazard ratios are appearing with +09

stcox ib4.grup age#age if procedure==1

I suspect because values are too small?

What can I do to get the readable output ?

And why are my values std error, z p value labelled as missing

Is it because I don't have much data?
I do have 400 observations, which I think should be enough.

[Andrew Musau]

You need to rescale your variables. If you have a variable like GDP, for example, express it in millions or billions of dollars instead of dollars.

[Clyde Schechter]

This is standard computer floating point notation. 5.54e+09 is how pretty much any computer program will display 5.54 x 10⁹. Now, conceivably you could request a different display format for these results using the -set cformat- command. (-help set cformat- if you are not familiar with this.)

But you have a much bigger problem here. There is almost nothing in the real world that should have a hazard ratio of 5,540,000,000. These numbers are all wildly unrealistic. I cannot tell from what is displayed which variable this number is the coefficient of. If it is coming from one of the levels of grup, then it suggests that in the reference category of grup, which your code sets as level 4, there is probably no failure event or failure events are very rare. You need to change the reference category to one that has a reasonable number of events occurring in the estimation sample. Maybe all of your categories of grup have only very rare failure events. If that's the case, it suggests that your data are simply inadequate for the purpose of investigating the failure rates among the categories of this variable--a much larger data set may be needed for the purpose.

If the numbers you show are coefficients of age or c.age#c.age, then it may be that you are using an inappropriate scale for age. If your age variable is currently in hours or days, change it to years or even decades to make these numbers more reasonable.

If you need more specific advice to chase this down, when posting back, use the -dataex- command to show example data and also show the full output of -stcox-, not just the numbers, so we can see which coefficients are affected and what kind of variables they are linked to.

Added: Crossed with #2.

[Tara Boyle]

Honestly, just fantastic advice. It’s not just an answer but such a logical explanation, i’ll check regarding my reference category and yes might be possible my category 4 , failures may be rare. I’ll switch it over for the ref category to be 1 - which has more failures.

will let you know how i got on

no the coefficients displays belong to the -grup- variable a linear score of 1-50

many thanks