2
$\begingroup$

I want to calculate $P(\neg H\mid I, \neg F)$,

where

$p(H) = p(\neg H) = 0.5$,

$p(I\mid H) = 0.8$, $p(\neg I\mid H) = 0.2$,

$p(I\mid\neg H) = 0.4$, $p(\neg I\mid\neg H) = 0.6$,

$p(F\mid H) = 0.3$, $p(\neg F\mid H) = 0.7$,

$p(F\mid\neg H) = 0.4$, $p(\neg F\mid\neg H) = 0.6$,

and $p(H\mid I) = 2/3$

However, I feel hard to make a formula with bayes rule.

It would be very helpful if somebody can derive it.

Answer:Thanks to Henry, I can understand how it derives. Let me add short answer to the post for whom who would face this problem like me.

$P(\neg H\mid I \cap \neg F) = \dfrac{P(\neg H \cap I \cap \neg F)}{P( I \cap \neg F)}$

$= \dfrac{P(\neg H \cap I) \times P(\neg H \cap \neg F)}{P(\neg H \cap I) \times P(\neg H \cap \neg F) + P(H \cap I) \times P(H \cap \neg F)}$ since $F \perp I$

$= 0.3$

$\endgroup$
  • 3
    $\begingroup$ You do not seem to know anything about the interaction between $I$ and $F$ $\endgroup$ – Henry Jan 1 '15 at 18:41
  • $\begingroup$ @Henry Does that mean the problem isn't solvable with just what zedoul gave us? $\endgroup$ – Zubin Mukerjee Jan 1 '15 at 18:43
  • $\begingroup$ @Zubin Mukerjee: I suspect not $\endgroup$ – Henry Jan 1 '15 at 18:47
1
$\begingroup$

You can calculate

  • $P(I \cap H)=0.4$
  • $P(\neg I \cap H)=0.1$
  • $P(I \cap \neg H)=0.2$
  • $P(\neg I \cap \neg H)=0.3$
  • $P(F \cap H)=0.15$
  • $P(\neg F \cap H)=0.35$
  • $P(F \cap \neg H)=0.2$
  • $P(\neg F \cap \neg H)=0.3$

and things based on this such as $P(I)=0.6$ and $P(\neg F)=0.65$.

But what you want to know is $ \dfrac{P(\neg H \cap I \cap \neg F)}{P( I \cap \neg F)}$ and the best you can do is put wide bounds on it.

For example, I think you can have $P(\neg H \cap I \cap \neg F)=0.2$ and $P(H \cap I \cap \neg F)=0.25$ to make your conditional probability $\frac49$.

I also think you can have $P(\neg H \cap I \cap \neg F)=0$ and $P(H \cap I \cap \neg F)=0.35$ to make your conditional probability $0$.

$\endgroup$
  • $\begingroup$ Thank you for your detailed solution. Let me ask one more thing. In order to calculate $P(¬H∩I∩¬F)$, first I can divide them into $P(¬H∩I∩¬F)=p(¬F|I∩¬H)p(I|¬H)p(¬H)$ However, how can I calculate $p(¬F|I∩¬H)$ in this setting? $\endgroup$ – zedoul Jan 1 '15 at 19:10
  • $\begingroup$ I do not think you can. The best you may be able to do is look at possibilities. $\endgroup$ – Henry Jan 1 '15 at 19:12
  • $\begingroup$ The value for $P(¬H|I,¬F)$ should be $0.3$ just with those setting. Anyway thank you for your answer! $\endgroup$ – zedoul Jan 1 '15 at 19:19
  • $\begingroup$ To get $0.3$ (which is within the range I gave) you have probably assumed that $I$ and $F$ are conditionally independent give $H$ or given $\neg H$. But the question does not suggest that you can assume this. $\endgroup$ – Henry Jan 1 '15 at 19:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.