Zoom Logo

Survival Analysis - Shared screen with speaker view
Will Lee
19:08
Where can we access the recording of the lecture?
Will Lee
19:20
Thank you!
Joni Ricks-Oddie
20:15
Congratulations Isabel!
Isabel Canette
20:28
Thanks :-)
Olivia Marie Silke
21:09
Is there a way to access the workshop materials?
Joni Ricks-Oddie
21:48
The materials will be available after the workshop. We will be uploading them here - https://statconsulting.uci.edu/workshop-materials/
Joni Ricks-Oddie
22:53
All previous workshop materials are also available on this webpage
Annie Ro
28:01
How is posttran different from transplant?
Isabel Canette
29:28
Posttran indicates if patient already had a transplant by time t. Transplant is just an indicator if the patient ever got a transplant, so it is not of interest in our analysis
Rupa Jose
31:03
sorry why again are there more obs or records than subjects...?
Isabel Canette
31:42
Some individuals have, for example data from before and from after having the transplant, or other variables might have changed
Isabel Canette
31:59
So this is why you have multiple records per individual
Rupa Jose
33:09
thanks!
Olivia Marie Silke
37:11
can you please repeat what the incidence risk is, I couldn't type my notes fast enough. Thanks!
Isabel Canette
37:51
That’s an interesting question for live answering, so we’ll answer it live.
Isabel Canette
38:05
Remember that the recording will be available, so you don’t have to take notes
Anton Palma
38:23
What's the interpretation for median survival or incidence rate by post-transplant? It seems odd that survival post-transplant is lower than pre-transplant
Annie Ro
38:40
is the test for trend testing the survival curves? Or just the events observed?
Rupa Jose
39:01
whats time at risk
Annie Ro
46:37
sorry not test for trend but the log likeihood
Annie Ro
47:06
yes this one.
Annie Ro
47:12
thanks
Kate Kirby
51:15
Re Anton’s question, it might be helpful to think of median survival time as median follow up instead—as that time statistic doesn’t provide any information about their vital status, only the amount of time they were in the study before transplant vs after transplant. Those receiving transplants are followed for some time after and may have a better survival rate (or maybe not) than those who did not receive transplants, but that’s not reflected in the median survival time.
Joni Ricks-Oddie
51:56
Thanks Kate!
Isabel Canette
53:46
Kate and Anton, it also seems an artifact of how the survival estimator is defined, you have to see the graph for that.
Joni Ricks-Oddie
55:28
Does the Stata documentation address the different methods for addressing ties and pros/cons of each method?
Isabel Canette
56:09
Yes, the documentation addresses that, but essentially is a trade-off between accuracy and computational efficiency
Anton Palma
56:10
Right, so it's comparing survival between post- vs. pre-transplant follow-up time, rather than survival between persons.
Joni Ricks-Oddie
56:45
Anton - That is what I was thinking as well
Isabel Canette
57:13
Yes, a patient might be in both groups.
Olivia Marie Silke
01:02:21
how do we address problems when the curves cross?
Isabel Canette
01:03:30
There are several ways to address when the PH doesn’t hold, let’s talk about this in the next break
Joni Ricks-Oddie
01:07:34
Would you ever use the tvc option with more then one variable?
Joni Ricks-Oddie
01:08:27
What do you mean by different functions of t?
Joni Ricks-Oddie
01:16:08
So theta is essentially the the variance around a random intercept like in a mixed model?
Isabel Canette
01:16:34
Yes, essentially, it is that
Joni Ricks-Oddie
01:18:41
In the first model there were also repeated measures. Why would you need to address the correlation between observations in the current example but not in the first example?
Joni Ricks-Oddie
01:33:16
A good paper on Competing risk models - Introduction to the Analysis of Survival Datain the Presence of Competing RisksPeter C. Austin, PhD; Douglas S. Lee, MD, PhD; Jason P. Fine, PhD
Yoko Narasaki
01:34:52
"sttcurve" is the best way for cometing risk analysis?---Is it also possible to use "stcompet" for competing risk analysis?
Yoko Narasaki
01:36:26
sorry I mean "stcrreg" not "sttcurve"
Yoko Narasaki
01:38:10
Thank you!
Joni Ricks-Oddie
01:41:28
Can you explain why a hazard would increase or decrease over time? How might a researcher know or identify that this is happening?
Joni Ricks-Oddie
01:50:00
Thanks
S. Mantravadi
01:57:51
Very informative! Would we be able to receive the recording or slide deck?
Joni Ricks-Oddie
01:58:41
Yes. They will be available on the Center for Statistical Consulting website here - https://statconsulting.uci.edu/workshop-materials/
Joni Ricks-Oddie
02:07:42
Since this is parametric model do you get a LR test for the random effect as in the frailty model?
Joni Ricks-Oddie
02:16:31
You still stset with gsem?
Gabriela Ortiz (StataCorp)
02:18:37
Yes, to fit a survival model with -gsem-, you would still -stset- your data, because -stset- will create that _t variable that is being used as the dependent variable.
Joni Ricks-Oddie
02:25:08
Ok thanks
Eunae Ju
02:28:01
Thank you !!