What is the Weibull Distribution Percentile Calculator?
This tool computes the percentile (also called the quantile or inverse CDF) of a two-parameter Weibull distribution. Given the shape parameter m, the scale parameter eta, and a cumulative probability, it returns the value x at which the distribution reaches that probability. It is a pure statistics tool that applies everywhere with no region-specific rules.
The Weibull distribution
The two-parameter Weibull distribution has a shape parameter \(m\) (sometimes written \(k\) or \(\beta\)) and a scale parameter \(\eta\) (sometimes written \(\alpha\) or \(\lambda\)), both strictly positive, with support \(x \ge 0\). Its lower cumulative distribution function is $$P = F(x) = 1 - \exp\left(-\left(\frac{x}{\eta}\right)^{m}\right).$$ The upper (survival) probability is $$Q = 1 - F(x) = \exp\left(-\left(\frac{x}{\eta}\right)^{m}\right),$$ so \(P + Q = 1\).
The quantile formula
Solving \(P = 1 - \exp\left(-\left(\frac{x}{\eta}\right)^{m}\right)\) for \(x\) gives the inverse CDF: $$x = \eta\left(-\ln\left(1-P\right)\right)^{\frac{1}{m}}.$$ If you supply an upper-tail probability \(Q\) instead, the calculator first converts it via \(P = 1 - Q\), so equivalently $$x = \eta\left(-\ln\left(Q\right)\right)^{\frac{1}{m}}.$$ The output \(x\) carries whatever unit the scale parameter represents (hours, cycles, etc.).
How to use it
Enter the shape parameter m and scale parameter eta. Choose whether your probability value is a lower cumulative probability P or an upper cumulative probability Q, then enter the probability strictly between 0 and 1. The result is the percentile x.
Worked example
With \(m = 2\), \(\eta = 1\), lower probability \(P = 0.5\): \(-\ln(1 - 0.5) = 0.693147\), and \(0.693147^{\frac{1}{2}} = 0.832555\), so $$x = 1 \times 0.832555 = 0.83255.$$ This is the median of the Weibull(2, 1) (Rayleigh) distribution.
Interpreting Weibull Percentiles and the Shape Parameter
The percentile \(x_P\) is the value below which a fraction \(P\) of the population falls, that is, \(F(x_P)=P\). In reliability work it is the time by which a fraction \(P\) of units are expected to have failed. For example, \(P=0.10\) gives the B10 life and \(P=0.5\) gives the median. Inverting the Weibull CDF gives:
$$x_P = \eta\left(-\ln(1-P)\right)^{1/m}$$
If you enter an upper (survival) probability \(Q\), the equivalent lower probability is \(P = 1-Q\).
Characteristic life
When \(x=\eta\), the CDF is \(1-e^{-1}\approx 0.632\), whatever the shape \(m\). So the scale parameter \(\eta\) (the characteristic life) is always the 63.2th percentile.
The shape parameter m
- m < 1: the failure rate decreases over time, which is typical of infant mortality or early defects.
- m = 1: the failure rate is constant. The Weibull reduces to the exponential distribution.
- m > 1: the failure rate increases over time, which is typical of wear-out and aging.
- m = 2: this is the Rayleigh distribution, where the failure rate increases linearly with time.
Scaling
For a fixed \(m\) and \(P\), the percentile scales linearly with \(\eta\). Doubling \(\eta\) doubles every percentile. The shape \(m\) controls how spread out the percentiles are: a larger \(m\) packs them more tightly around \(\eta\).
Common Reliability Percentiles (B-Life) Reference
Each percentile is \(x = \eta \cdot f^{1/m}\), where \(f=-\ln(1-P)\) is the factor in the table. The last column gives \(x/\eta\) for the Rayleigh case \(m=2\), where the exponent \(1/m\) is simply a square root. For \(m=1\), \(x/\eta\) equals the factor itself.
| Label | P (cumulative probability) | Factor \(-\ln(1-P)\) | \(x/\eta\) at \(m=2\) |
|---|---|---|---|
| B1 | 0.01 | 0.01005 | 0.1002 |
| B5 | 0.05 | 0.05129 | 0.2265 |
| B10 | 0.10 | 0.10536 | 0.3246 |
| B50 (median) | 0.50 | 0.69315 | 0.8326 |
| Characteristic life | 0.632121 | 1.00000 | 1.0000 |
| B90 | 0.90 | 2.30259 | 1.5174 |
| B99 | 0.99 | 4.60517 | 2.1460 |
Example: for \(\eta=1000\) and \(m=2\), the B10 life is \(x = 1000 \times 0.10536^{1/2} \approx 324.6\).
Note that the factor does not depend on the shape. The shape only enters through the exponent \(1/m\).
FAQ
What if I have the reliability (survival) probability? That is the upper probability \(Q\); select "Upper cumulative probability Q" and enter it directly.
Why must the probability be strictly between 0 and 1? As \(P\) approaches 1 the percentile tends to infinity, and at \(P = 0\) it is 0; values at or beyond the boundaries make the logarithm undefined.
Can the result be negative? No. The Weibull support is \(x \ge 0\), so the percentile is always non-negative.