how to test a hypothesis when given population mean and standard deviation is known?

680 Views Asked by At

the mean weight of a 6 year old child is 49.3 lbs. another study indicated that in a sample of 196 six year olds the mean weight was 51.5 lbs. Assume a population standard deviation of 14. use the significance level 0.01 to determine z score and corresponding p value.

I've been stuck on this question for quite some time now, can anyone help me out?

1

There are 1 best solutions below

0
On

The null hypothesis is that the mean weight is 49.3 lbs, with 14 lbs. standard deviation. The question is, what is the probability that you observed a mean of 51.5 lbs. in a sample of 196 children, if the true mean and standard deviation followed the null hypothesis?

We'll assume that a great many random factors are involved in the weight of any particular child, and that the sampling of children was random, and that their weights are independent; then by appealing to the central limit theorem (and observation), we'd expect the weights of children to be distributed as a normal distribtuion.

Let $X_n\sim\mathcal{N}(\mu,\sigma^2)$ be independent, indentically distributed random variables, for $n=1,2,\ldots N$, each representing the weight of a child. $N=196$, $\mu=49.3$, and $\sigma=14$. What we observed was the sample mean

$$ \hat\mu = \frac{1}{N}\sum_{n=1}^N X_n $$

By the properties of normal random variables, the sample mean is also distributed as a normal, with $\hat\mu\sim\mathcal{N}(\mu,\sigma^2/N)$. Now our question is, what were the chances that we observed a sample mean different than the true mean? How often could we have expected to see this value, or a more extreme value? This is the definition of the $p$-value. $$ p = \Pr[\hat\mu\geq51.5] = \Pr[\mathcal{N}(\mu,\sigma^2/N)\geq51.5] = $$ $$ \Pr\left[\mathcal{N}(0,1)\geq \frac{51.5-49.3}{14/\sqrt{196}}\right] = 1-\Phi(2.2) \approx 1-0.9861 = 0.0139 = 1.39\% $$ So if the null hypothesis was true to begin with, there is slightly more than a $1\%$ chance that we'd observe a sample mean of $51.5$ in a sample of $196$. That's a rare-sounding outcome; I'll leave it to you to figure out if this is significant at the $1\%$ level, and what the z-score is above.