MLE $\hat{\theta} = 0$ for a binomial distributed CDF?

57 Views Asked by At

Suppose 100 seeds were planted and it is recorded that $x < 11$ seeds germinated. Assuming a binomial model we obtain the following likelihood: binomial image

I believe the curve can be calculated with the following r-code:

curve(pbinom(11, 100, y), xname = "y")

We have that:

$$L(\theta) = \sum_{x=0}^{10} \binom{n}{x}\theta^x(1-\theta)^{n-x}$$

The MLE is $\hat{\theta}=0$ However, how do I calculate this MLE?

I have tried:

$$L(\theta;x) = \prod_{i=1}^{10}f(x_i;\theta) = \prod_{i=1}^{10}\sum_{x=0}^{10}\binom{n}{x}\theta^x(1-\theta)^{n-x} \\ \implies \sum_{x=0}^{10}\binom{n}{\prod_{i=1}^{10}x}\theta^{\sum_{i=1}^{10}x_i}(1-\theta)^{10n-\sum_{i=1}^nx_i}$$

Then taking the logarithm: $$\log L(\theta;x) = \log \left(\sum_{x=0}^{10}\binom{n}{\prod_{i=1}^{10}x}\right)+\sum_{i=1}^{10}x_i\log \left(\sum_{x=0}^{10}\theta\right)+(10n-\sum_{i=1}^{10}x_i) \log \left(\sum_{x=0}^{10}(1-\theta)\right)$$

We know that the MLE of a binomial distribution is given as $\hat{\theta} = \frac{x}{n}$ so I cannot understand how the MLE for the likelihood is 0. I suppose my way of indexing summation and products is wrong here?

Taken from: Pawitan, Y. 2013. In All Likelihood: Statistical Modelling and Inference using Likelihood, pg 40