An experiment is designed to test the potency of a drug on 40 rats. Previous animal studies have shown that a $10$-mg dose is lethal $10$% of the times within the first $4$ hours.

What is the probability that between $2$ and $8$ rats die during the experiment due to the drug?

My attempt: Let $X$ be the number of rats that die in the first 4 hours. $$P(2 \leq x \leq8) = \sum_{x=2}^{8} \binom{40}{x} (0.1)^{x} (0.9)^{40-x}$$

To make the drug more potent, the company came with a new formula. This reduced the chances of a new drug being lethal to $1$%. The new drug is administered to $10000$ rats. Approximate the exact probability that $5$ rats die.

My attempt:

$$P(x=5) = \binom{10000}{5} (0.01)^{5} (0.99)^{10000-5}$$

Am I on the right path?

  • 2
    $\begingroup$ The wording in the second question is odd. I think it should be "the probability that exactly 5 rats die." However you are right. $\endgroup$ – Shogun Sep 13 '18 at 19:16
  • $\begingroup$ 2nd question, you mean $5$ instead of $x$, right ? $\endgroup$ – Ahmad Bazzi Sep 13 '18 at 19:25
  • $\begingroup$ @AhmadBazzi, yes! I made the necessary corrections. But is the set-up right? $\endgroup$ – Lady Sep 13 '18 at 19:31
  • $\begingroup$ @RHowe, any idea what the exact number is? $\endgroup$ – Lady Sep 13 '18 at 20:28
  • $\begingroup$ @Lady I've added an answer $\endgroup$ – Shogun Sep 13 '18 at 21:13

In order to calculate the first statement

$$ P(2 \leq X \leq 8) = \sum_{x=2}^{8} \binom{40}{x} \big(0.1\big)^{x} \big(0.9\big)^{40-x} \tag{1}$$

We can utilize some small python script as it is kind of hard by hand.

from scipy import stats

# number of rats
# lethal dose probability
# upper bounds
kmax = 8
kmin = 1
# cdf summations
y1 = stats.binom.cdf(kmin, n, p)
y2 = stats.binom.cdf(kmax,n,p)
# differences
y = y2-y1
Out[13]: 0.9040309985710783

Note the binom.cdf command is

$$ F(x|n,p) = \sum_{i=0}^{n} \binom{n}{i} \big(p)^{i} \big(1-p\big)^{n-i} \tag{2}$$

So we to set our limits

$$ I_{1} = F(x|1,0.1) = \sum_{i=0}^{1} \binom{1}{i} \big(0.1)^{1} \big(0.9 \big)^{1-i} \tag{3}$$

$$ I_{2} F(x|8,0.1) = \sum_{i=0}^{8} \binom{8}{i} \big(0.1)^{8} \big(0.9 \big)^{8-i} \tag{4}$$ then we find

$$ I = I_{2} - I_{1} \tag{5} $$ there is a 90% chance $2$ - $8$ rats die.

ok..the second one

$$ P(X=5) = \binom{10000}{5} \big(0.01\big)^{5} \big(0.99\big)^{9995} \tag{6}$$

from scipy import stats

# number of rats
# lethal dose probability
# exact number 
k = 5
# pmf 
y = stats.binom.pmf(k, n, p)
Out[8]: 1.968586020940221e-36

Wow, startling difference.. I don't think you can do these by hand. You can make rough approximations.

In the last equation you get some rough terms...

$$ P(X=5) = \frac{10000!}{5!9995!} \frac{1}{100^{5}} \frac{99^{9995}}{100^{9995}} \tag{7}$$ Various things will drop out

$$ P(X=5) = \frac{10000\cdot 9999 \cdot 9998\cdot 9997\cdot 9996 }{120} \frac{1}{100^{2} \cdot 10^{3}} \frac{99^{9995}}{100^{9995}} \tag{8}$$

$$ P(X=5) = \frac{ 9999 \cdot 9998\cdot 9997\cdot 9996 }{120} \frac{1}{ 10^{3}} \frac{99^{9995}}{100^{9995}} \tag{9}$$

It just becomes a really small number..this is the exact reason calculators were invented.

  • $\begingroup$ quite startling difference. An exact probability that 5 rats would die would be 1.9686X10^-36. Would that be an exact number? $\endgroup$ – Lady Sep 13 '18 at 21:22
  • $\begingroup$ I think I messed up this first calculation, I think it is actually 20% higher but it is mainly because of how many rats there are and the difference in probability. $\endgroup$ – Shogun Sep 13 '18 at 21:24
  • $\begingroup$ I did the computation by hand for the first one and had 0.904030995. Which is 90% chance that 2 to 8 rats would die. $\endgroup$ – Lady Sep 13 '18 at 21:31
  • $\begingroup$ yep, it's fixed..momentary lapse of judgment. $\endgroup$ – Shogun Sep 13 '18 at 21:33
  • 1
    $\begingroup$ it's effectively zero $\endgroup$ – Shogun Sep 13 '18 at 21:48

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.