Admin 08 Jun 2026 00:08

admin #general

Population, Sample, and Statistical Inference

Statistical inference is the process of drawing conclusions about a large groupcalled a populationbased on the analysis of a smaller set of datacalled a sample. The core ideas of probability, sampling, and inference are intertwined and provide the foundation for virtually every scientific discipline that uses data.

1. What Is a Population?

A population comprises all possible observations that share a set of characteristics. The definition can be tangible, such as all residents of New York City, or abstract, such as the set of all possible outcomes when rolling a fair die. Because it is usually impossible (or impractical) to collect data from every member, researchers treat the population as a conceptual entity with specific parameters:

Mean () the average value of a numeric characteristic across the whole population.
Variance () the spread of the population values around the mean.
Proportion (p) the fraction of the population that possesses a particular attribute (e.g., the proportion of voters who support a candidate).

These parameters are fixed, but they are typically unknown, which is why we turn to sampling.

2. What Is a Sample?

A sample is a subset of the population that is selected for measurement. When the sampling method is sound, the sample will represent the population, meaning the samples statistics (sample mean, sample variance, sample proportion) are close to the corresponding population parameters.

2.1 Types of Sampling Designs

Simple Random Sampling (SRS) Every individual has an equal chance of being selected. This is the gold standard because it eliminates systematic bias.
Stratified Sampling The population is divided into mutually exclusive subgroups (strata) such as age or income, and a random sample is taken from each stratum. This technique improves precision when the strata differ substantially.
Cluster Sampling Whole groups (clusters) are sampled, often for logistical convenience. For example, choosing entire schools rather than individual students.
Systematic Sampling After a random start, every kth element is selected. It is simple but can be vulnerable to hidden patterns.

2.2 Sample Size

Sample size (n) directly influences the accuracy of inference. Larger samples reduce random error, as reflected by smaller standard errors. However, larger samples cost more time and resources, so researchers balance precision with feasibility. Formulas for estimating required n often involve desired confidence level, margin of error, and an estimate of the population variance or proportion.

3. Probability Foundations

Probability quantifies uncertainty. In the context of sampling, it describes the likelihood of observing particular sample outcomes if we were to repeat the sampling process many times (the socalled sampling distribution). Two concepts are central:

3.1 Random Variables

A random variable (RV) associates a numeric value with each possible outcome of a random experiment. For example, X could denote the weight of a randomly selected adult. When we talk about the sampling distribution of the mean, we treat the sample mean \(\bar{X}\) itself as a random variable.

3.2 Central Limit Theorem (CLT)

The CLT states that, for a sufficiently large sample size, the sampling distribution of the sample mean is approximately normal, regardless of the shape of the underlying population distribution. Formally:

\(\displaystyle \bar{X}\; \approx\; N\!\Big(\mu,\ \frac{\sigma^{2}}{n}\Big)\)

This result justifies using normalbased confidence intervals and hypothesis tests even when the original data are not normally distributed.

4. Statistical Inference: From Sample to Population

4.1 Point Estimation

A point estimator is a single number derived from the sample that serves as a best guess for an unknown population parameter. Common estimators include:

Sample mean \(\bar{x}\) for estimating .
Sample proportion \(\hat{p}\) for estimating p.
Sample variance s for estimating .

4.2 Interval Estimation

Because a point estimate is subject to sampling variability, we usually report an interval that is likely to contain the true parameter with a specified confidence level (e.g., 95%).

Confidence interval for a mean ( known):

\(\displaystyle \bar{x}\ \pm\ z_{\alpha/2}\,\frac{\sigma}{\sqrt{n}}\)

Confidence interval for a mean ( unknown):

\(\displaystyle \bar{x}\ \pm\ t_{\alpha/2,\,df=n-1}\,\frac{s}{\sqrt{n}}\)

Confidence interval for a proportion:

\(\displaystyle \hat{p}\ \pm\ z_{\alpha/2}\,\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}\)

4.3 Hypothesis Testing

Hypothesis testing evaluates whether observed data provide sufficient evidence to reject a specific claim about a population parameter. The steps are:

State null (H) and alternative (H) hypotheses.
Choose a significance level (commonly 0.05).
Compute a test statistic (e.g., z, t, ).
Determine the pvalue or critical region.
Make a decision: reject H if p , otherwise fail to reject.

Typical tests include the onesample ttest for means, the twosample ttest for comparing two groups, and the test for categorical data.

5. Common Sources of Error

5.1 Random Error

Random error arises from the natural variability inherent in sampling. It is quantified by the standard error and can be reduced by increasing the sample size.

5.2 Systematic Error (Bias)

Bias occurs when the sampling design or measurement procedure consistently deviates from the truth. Examples include:

Selection bias certain subpopulations are over or underrepresented.
Measurement bias instruments give inaccurate readings.
Nonresponse bias individuals who do not respond differ from those who do.

5.3 Model Misspecification

When the statistical model (e.g., assuming normality) does not reflect the data-generating process, estimators may be biased or have inflated variance. Diagnostic plots and goodnessoffit tests help detect such problems.

6. Practical Example

Suppose a public health agency wants to estimate the average daily sodium intake of adults in a city. The population of interest is all adults (500,000). The agency draws a simple random sample of n = 400 adults and records their sodium consumption (in mg). The sample mean is \(\bar{x}=2{,}300\) mg and the sample standard deviation is s = 600 mg.

6.1 Point Estimate

The best guess for the city's average intake is 2,300 mg.

6.2 95% Confidence Interval

Because is unknown, we use the tdistribution with df = 399. The critical value \(t_{0.025,399}\) 1.97.

\(\displaystyle \text{CI}=2{,}300 \pm 1.97\frac{600}{\sqrt{400}}=2{,}300 \pm 1.97(30)=2{,}300 \pm 59\)

Thus, we are 95% confident that the true mean sodium intake lies between 2,241 mg and 2,359 mg.

6.3 Hypothesis Test

Public health guidelines recommend a maximum of 2,300 mg. To test whether the city's average exceeds this limit, set:

H: = 2,300 mg
H: > 2,300 mg

The test statistic is

\(\displaystyle t=\frac{\bar{x}-2{,}300}{s/\sqrt{n}}=\frac{2{,}300-2{,}300}{30}=0\)

With a pvalue of 0.5, we fail to reject H. The data do not provide evidence that the average intake exceeds the recommended level.

7. Key Takeaways

The population is the full set of entities we wish to understand; the sample is the tractable subset we actually observe.
Probability theory underpins the behavior of sample statistics and enables us to quantify uncertainty.
Point estimates give a single best guess, while confidence intervals convey the range of plausible values.
Hypothesis testing provides a formal framework for evaluating specific claims about population parameters.
Careful sampling design and awareness of potential biases are essential for reliable inference.