Hi everyone,
I want to compare the following outcome between two independent groups:
anesthesia time
time to surgical readiness
surgical time,
emergence time,
and total operating room time
I am wondering should I use survival analyze or independent t test? I know survival is for time to event data and censored data, and here as you see my outcomes variables is related to time. however I think it is possible to compare the mean of these outcome using t -test
can I say if I have a censored data I use survival analyze, otherwise I will use the simplest way which is t-test to do comparison?
Thanks.
No. Waiting times (time-to-event data, survival times) have a skewed distribution. Parametric analyses should be based on appropriate distributions like Weibull, Gamma or log-normal, but not on the normal distribution. You might use generalized models and approximate likelihood ratio tests.
you should provide the data distribution to allow your readers to reply correctly.
read again JW answer
Bahar Ysr wrote: I know survival is for time to event data and censored data, and here as you see my outcomes variables is related to time. (Emphasis added.)
Are you suggesting that time-to-event (TTE) analysis can only be used if there is some censoring?
Thank you for your answer. I believe survival analysis can also be used without censored data.
What I was confused is "time to surgical readiness" is the time to event variable, so I will go with survival. no issue with this outcome variable.
But what about the other variables? I can see some papers that used the Man Witney test to make the comparison such as total operating room, but based on the explanation from Joshen W, I think it is not suitable to go with this method. it is better to use survival or GLM. Am I correct?
anesthesia time
time to surgical readiness
surgical time,
emergence time,
and total operating room time
Thanks
But for all of those variables, there is a start time and a finish time, isn't there? Take surgical time, for example. The surgery begins at one point in time and ends at another. And the event for a TTE model would be completion of the surgery. Or have I misunderstood something?
With 5 outcome variables, another issue to think about is multiplicity. The two 2005 Lancet articles by Schulz & Grimes do a great job of discussing the issues, IMO. HTH.
Article Schulz K, Grimes DMultiplicity in randomised trials I: endpo...
Article Multiplicity in randomised trials II: Subgroup and interim analyses
Thanks for your answer. Based on your explanation, the reason those papers used the Wilcoxson test is that there is no censorship. All instances should have start and end date to use this type of analyze.
I haven't seen my data yet to be sure about censorship.
So, I am eligible to use Wilcox if there is no censorship. The only issue with that would be multiplicity, as I have 5 outcome variables.
The Wilcox test will test a hypothesis you are likely not interested in. What is the hypothesis you want to test and what is the hypothesis tested by the Wilcox test?
provide the data and I (but others can) will show you the analysis
Cheers
Hello Bahar Ysr My idea is that in survival analysis, we are faced with three basic elements and words that constitute this type of statistical analysis. These three words are probability, event and time. It seems to me that your study lacks these three essential parts and just because there is something called time does not make you want to use survival analysis. in this study, I don't see anything called an event (happening to some people and not to others).
The Wilcoxson or Mann-Witney- u test will be used to compare the mean of two continuous variables between two groups to see if there is any significant difference.
Here, the hypothesis is that, for instance, Anesthesia time and time to surgical readiness are expected to be longer for lobectomies compared to laparoscopic colectomies.
This is not what the Wilcoxson or Mann-Witney- u test does (only under some additional assumptions that are not sensible for time-to-event data).
The "additional assumptions" Jochen Wilhelm mentioned earlier are described as the "pure shift" model by Fagerland & Sandvik (2009). I have attached an excerpt from their article's introduction.
Article The Wilcoxon-Mann-Whitney test under scrutiny
- Badges
- Science topic
- Similar topics
- Psychology