Mathwords logoMathwords

Survival Function — Definition, Formula & Example

The survival function is the probability that something, such as a patient, a machine part, or a light bulb, is still working or alive after time t. It starts at 1 and falls toward 0 as time passes.

For a nonnegative random variable TT with cumulative distribution function FF, the survival function is S(t)=P(T>t)=1−F(t)S(t) = P(T > t) = 1 - F(t) for t≥0t \ge 0. SS is non-increasing and right-continuous, with lim⁡t→∞S(t)=0\lim_{t \to \infty} S(t) = 0. If P(T=0)=0P(T = 0) = 0, then S(0)=1S(0) = 1. When TT has a density ff, S(t)=∫t∞f(u) duS(t) = \int_t^{\infty} f(u)\,du and f(t)=−S′(t)f(t) = -S'(t).

Key Formula

S(t)=P(T>t)=1−F(t)=∫t∞f(u) duS(t) = P(T > t) = 1 - F(t) = \int_t^{\infty} f(u)\,du
Where:
  • TT = Random lifetime, the time until failure or death
  • tt = A fixed time, with t ≥ 0
  • F(t)F(t) = Cumulative distribution function, P(T ≤ t)
  • f(u)f(u) = Probability density function of T

Worked Example

Problem: A water pump's lifetime T, in years, has survival function S(t)=e−(t/10)2S(t) = e^{-(t/10)^2}. Find the probability that the pump lasts more than 5 years and the probability that it lasts more than 10 years. Then find the probability that a pump still running at year 5 is still running at year 10.
Step 1: Evaluate the survival function at t = 5.
S(5)=e−(5/10)2=e−0.25≈0.7788S(5) = e^{-(5/10)^2} = e^{-0.25} \approx 0.7788
Step 2: Evaluate the survival function at t = 10.
S(10)=e−(10/10)2=e−1≈0.3679S(10) = e^{-(10/10)^2} = e^{-1} \approx 0.3679
Step 3: Lasting past 10 years already implies lasting past 5 years. So the conditional probability is a ratio of survival values.
P(T>10∣T>5)=S(10)S(5)=e−1e−0.25=e−0.75≈0.4724P(T > 10 \mid T > 5) = \frac{S(10)}{S(5)} = \frac{e^{-1}}{e^{-0.25}} = e^{-0.75} \approx 0.4724
Answer: S(5)≈0.7788S(5) \approx 0.7788 and S(10)≈0.3679S(10) \approx 0.3679. A pump that is still running at year 5 has about a 47.24% chance of reaching year 10.

Visualization

Why It Matters

Engineers use survival functions to set warranty periods and maintenance schedules. Medical researchers use them to compare treatments, often estimating S(t) from patient data with the Kaplan–Meier method. The hazard rate, the main tool for modeling failure risk, is defined as h(t)=f(t)/S(t)h(t) = f(t)/S(t).

Common Mistakes

Mistake: Assuming a unit that has survived 5 years is as good as new, so its chance of lasting 5 more years is just S(5) ≈ 0.7788.
Correction: Among continuous lifetime distributions, that shortcut works only for the exponential distribution, which is memoryless. For this pump the conditional chance is S(10)/S(5) ≈ 0.4724, because older pumps wear out faster. In general, compute P(T > t + s | T > t) as S(t + s)/S(t).

Related Terms