The course consists of concepts of sampling, testing hypothesis, parametric and non parametric tests, correlation and regression, experimental designs and stochastic processes.
STA215Semester III3 credit hours6 units71 solved questions939 study cardsFull study modes
The course consists of concepts of sampling, testing hypothesis, parametric and non parametric tests, correlation and regression, experimental designs and stochastic processes.
The free sample is open to everyone. Sign in for 2 free weeks of every answer, flashcards and progress.
This chapter moves from describing a single sample to reasoning about the population it came from, building the sampling distributions, point and interval estimates, and sample size formulas that all later inferential techniques rely on.
A sampling distribution is the probability distribution of a statistic (such as a mean, proportion, or variance) computed over every possible sample of a fixed size $n$ drawn from a population. It is distinct from the population distribution, which describes individual values in the population, and from a single sample's own data distribution. Because a statistic varies from sample to sample purely by chance, it behaves like a random variable with its own mean, variance, and shape, and the standard error is the name given to the standard deviation of a sampling distribution: it measures how much the statistic is expected to fluctuate from one sample to the next. Sampling distributions are the bridge between a single observed sample and a statement about the unknown population parameter, since every confidence interval and every hypothesis test is built directly on one.
If a population has mean $\mu$ and variance $\sigma^2$, the sampling distribution of the sample mean $\bar{x}$ based on samples of size $n$ has $$E(\bar{x}) = \mu, \qquad V(\bar{x}) = \frac{\sigma^2}{n} \ \text{(infinite population, or sampling with replacement)}$$ so its standard error is $\sigma_{\bar{x}} = \dfrac{\sigma}{\sqrt{n}}$. When the population is finite of size $N$ and sampling is without replacement, a finite population correction applies: $$\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}}\sqrt{\frac{N-n}{N-1}}$$ which shrinks the standard error as the sample becomes a larger fraction of the population. Similarly, for a population proportion $P$ (with $Q = 1-P$), the sampling distribution of the sample proportion $\hat{p} = \dfrac{X}{n}$, where $X$ is the number of successes in the sample, has $$E(\hat{p}) = P, \qquad V(\hat{p}) = \frac{PQ}{n}, \qquad \sigma_{\hat{p}} = \sqrt{\frac{PQ}{n}}$$ again with a finite population correction $\sqrt{\dfrac{N-n}{N-1}}$ multiplying the standard error when sampling without replacement from a finite population. In both cases the mean of the sampling distribution equals the population parameter, meaning $\bar{x}$ and $\hat{p}$ are unbiased, and the spread of the sampling distribution shrinks as $n$ grows.
The Central Limit Theorem (CLT) states that if random samples of size $n$ are drawn from any population with finite mean $\mu$ and finite variance $\sigma^2$, then as $n$ becomes large (a common working rule is $n \ge 30$), the sampling distribution of the sample mean $\bar{x}$ approaches a normal distribution $N!\left(\mu, \dfrac{\sigma^2}{n}\right)$, regardless of the shape of the parent population. This is what allows the standard normal table to be used for sample-mean problems even when the underlying population is skewed or unknown, and it is the theoretical justification for every large-sample confidence interval and test built later in this chapter.
Worked example. Response times logged from a large web service have population mean $\mu = 50$ ms and standard deviation $\sigma = 10$ ms. For a random sample of $n = 64$ requests, the standard error of the sample mean is $$\sigma_{\bar{x}} = \frac{10}{\sqrt{64}} = 1.25$$ By the CLT, $\bar{x} \sim N(50, 1.25^2)$ closely enough for the sample size, so the probability that the sample mean exceeds 52 ms is found by standardizing: $$Z = \frac{52-50}{1.25} = 1.6, \qquad P(\bar{x} > 52) = 1 - \Phi(1.6) = 1 - 0.9452 = 0.0548$$ so only about 5.5% of samples of size 64 would show a mean response time above 52 ms, even though individual request times themselves may be far more variable than this.
Statistics II for TU BSc CSIT is covered here in full, from syllabus notes to solved past questions. This subject has 6 chapters mapped to the official syllabus, and 71 past exam questions with written model answers. The opening of every note and answer is free, and the most-searched questions are free in full. A free account gets 2 free weeks of the whole program plus flashcards and progress tracking so weak topics come back before the exam.
This TU BSc CSIT Statistics II guide covers Sampling Distribution and Estimation, Testing of hypothesis, Non parametric test, Multiple correlation and regression, Design of experiment, Stochastic Process, across 6 chapters. It includes model answers to 71 past exam questions, plus flashcards for quick revision.
Statistics II has 6 chapters mapped to the official syllabus.
Yes. Each of the 71 past questions comes with a written model answer; the most-searched ones are free in full, the rest open with a free account's 2 free weeks or a pass.
Free to start. The opening of every note and answer is free, the most-searched questions are free in full, and a free account gets 2 free weeks of the whole program plus flashcards and progress tracking. After that a program pass keeps everything open.
Based on the official syllabus.
This chapter moves from describing a single sample to reasoning about the population it came from, building the sampling distributions, point and interval estimates, and sample size formulas that all later inferential techniques rely on.
A sampling distribution is the probability distribution of a statistic (such as a mean, proportion, or variance) computed over every possible sample of a fixed size drawn from a population. It is distinct from the population distribution, which describes individual values in the population, and from a single sample's own data distribution. Because a statistic varies from sample to sample purely by chance, it behaves like a random variable with its own mean, variance, and shape, and the standard error is the name given to the standard deviation of a sampling distribution: it measures how much the statistic is expected to fluctuate from one sample to the next. Sampling distributions are the bridge between a single observed sample and a statement about the unknown population parameter, since every confidence interval and every hypothesis test is built directly on one.
If a population has mean and variance , the sampling distribution of the sample mean based on samples of size has so its standard error is . When the population is finite of size and sampling is without replacement, a finite population correction applies: which shrinks the standard error as the sample becomes a larger fraction of the population. Similarly, for a population proportion (with ), the sampling distribution of the sample proportion , where is the number of successes in the sample, has again with a finite population correction multiplying the standard error when sampling without replacement from a finite population. In both cases the mean of the sampling distribution equals the population parameter, meaning and are unbiased, and the spread of the sampling distribution shrinks as grows.
The Central Limit Theorem (CLT) states that if random samples of size are drawn from any population with finite mean and finite variance , then as becomes large (a common working rule is ), the sampling distribution of the sample mean approaches a normal distribution , regardless of the shape of the parent population. This is what allows the standard normal table to be used for sample-mean problems even when the underlying population is skewed or unknown, and it is the theoretical justification for every large-sample confidence interval and test built later in this chapter.
Worked example. Response times logged from a large web service have population mean ms and standard deviation ms. For a random sample of requests, the standard error of the sample mean is By the CLT, closely enough for the sample size, so the probability that the sample mean exceeds 52 ms is found by standardizing: so only about 5.5% of samples of size 64 would show a mean response time above 52 ms, even though individual request times themselves may be far more variable than this.