Updating after each test result

A patient takes the same test several times. Update the probability of illness after each result, using each answer as the next prior.

What to build

A program that applies Bayes' theorem once per test result.

Requirements

  • The test has a sensitivity (P(positive | sick)) and a specificity (P(negative | healthy)); results are independent given the patient's state.
  • For each result print after R: P with R the result (+ or -) and P the probability of illness so far, to 4 decimal places.

Input

Line 1: the prior, the sensitivity and the specificity. Line 2: the results, + or -, separated by spaces.

Output

One line per result.

Constraints

1 to 20 results; the numbers are decimals from 0 to 1, and no denominator is 0.

Example

Input:

0.01 0.99 0.95
+ +

Output:

after +: 0.1667
after +: 0.7984

Why it works

Each result is evidence that Bayes' theorem combines with the current belief, so the posterior after one test is the prior for the next, and repeated positives quickly overcome a low base rate.

Examples

Example 1 · Input
0.01 0.99 0.95
+ +
Expected output
after +: 0.1667
after +: 0.7984
Example 2 · Input
0.5 0.8 0.9
- +
Expected output
after -: 0.1818
after +: 0.6400

Prompts inside input() are ignored when checking, so input("n: ") is fine.

Hints

Hint 1

After a positive: p × sens ÷ (p × sens + (1 − p) × (1 − spec)).

Hint 2

After a negative: p × (1 − sens) ÷ (p × (1 − sens) + (1 − p) × spec). The new p is the prior for the next result.

Solution

Show one correct solution (try it yourself first)
p, sens, spec = map(float, input().split())
for r in input().split():
    if r == "+":
        yes, no = p * sens, (1 - p) * (1 - spec)
    else:
        yes, no = p * (1 - sens), (1 - p) * spec
    p = yes / (yes + no)
    print(f"after {r}: {p:.4f}")

Lesson: Bayes' theorem →