topic Re: Survival analysis - how to calculate survival probability for a specific time in Discussions

Survival analysis - how to calculate survival probability for a specific time

hillaryvoet — Sat, 01 Jul 2023 20:15:49 GMT

When using the Survival platform, one of the options listed on the output is "Estimate survival probability", but this option is blacked out and cannot be applied. Why is this?

Re: Survival analysis - how to calculate survival probability for a specific time

txnelson — Sun, 02 Jul 2023 04:15:01 GMT

The documentation in the "Reliability and Survival Methods" document specifies

Estimate Survival Probability

Estimates the failure and survival probabilities for the given
time values. Specify effect values and one or more time values. JMP then calculates the
survival and failure probabilities with 95% confidence limits for all possible combinations
of the entries

There is an example using Parametric Survival Fit platform and Estimating the Survival Probabilities in the chapter "Fit Parametric Survival". I suggest you look there for the information you are asking about.

Re: Survival analysis - how to calculate survival probability for a specific time

hillaryvoet — Sun, 02 Jul 2023 05:47:28 GMT

Thank you. I'll look into it.

Re: Survival analysis - how to calculate survival probability for a specific time

Mark_Bailey — Mon, 03 Jul 2023 19:46:39 GMT

The legacy Survival platform estimates survival using the non-parametric Kaplan-Meier product moment calculation. You obtain survival from the plot or open the Combined outline to make it easier.

You can also fit a survival distribution using commands from the platform red triangle menu. This fit enables the Estimate survival probability command in the same menu.

I suggest using the modern Life Distribution platform instead of the legacy one.

Re: Survival analysis - how to calculate survival probability for a specific time

hillaryvoet — Mon, 03 Jul 2023 18:38:53 GMT

Thank you. I want to estimate survival probabilities without assuming a distribution, i.e. by the Kaplan Meier method. In the Fit options the nonparametric option does not appear. The survival probabilities are given in the output for all survival values which were observed. If, for example, I want a survival probability for 5 years, but no case in the dataset survived for exactly 5 years, I need to interpolate "manually".

I didn't see an option to estimate specific survival probabilities on the Life Distribution platform.

Re: Survival analysis - how to calculate survival probability for a specific time

Mark_Bailey — Mon, 03 Jul 2023 19:56:24 GMT

The non-parametric estimate of the survival probability is constant over a given interval. For example, here is an analysis in which I want to estimate the survival at 132.

Open the Combined outline in the Survival platform and find 132 in the table.

The estimate is 0.1 over the interval [128,134].

The Nonparametric Estimate in Life Distribution provides the same information in the last column: 1 - failure = survival = 1 - 0.9 = 0.1. Many analysts prefer the mid-point estimate to the original K-M estimate.

Re: Survival analysis - how to calculate survival probability for a specific time

hillaryvoet — Tue, 04 Jul 2023 05:48:30 GMT

Got it! Thank you!