This method uses intact groups as the primary units of selection, then collects data from members within those chosen groups. You start by identifying clusters that already exist, like classrooms or neighborhoods, then you pick some of those clusters and sample within them. It’s a practical approach when it’s easier to work with groups than with individuals scattered across a broad area, and it can save time and resources while still giving representative data if the clusters are similar to the whole population. Why this fits your scenario: selecting entire clusters (classrooms) and then sampling inside those clusters is exactly how this method operates—you don’t randomize individuals from the entire population at once, but rather focus on whole groups and then sample inside them. In contrast, simple random sampling would involve picking individuals at random from the entire population without using groups; stratified sampling would divide the population into strata and sample from each stratum to ensure representation across groups; systematic sampling would select every kth individual from a list.

Handling missing data in quantitative analyses hinges on identifying why data are missing and what that means for the results. The key idea is to distinguish patterns of missingness: MCAR (missing completely at random), MAR (missing at random), and MNAR (missing not at random). Each pattern points to different reasonable strategies. If data are MCAR, deleting cases with missing values can be unbiased, though it still reduces power. If data are MAR, imputation or model-based approaches can yield unbiased estimates by using observed information to estimate missing values. If data are MNAR, standard methods may still be biased, and specialized models or sensitivity analyses are needed. Because we don’t know the pattern beforehand, the prudent move is to assess the missingness pattern first and then decide among deletion, imputation (such as multiple imputation, which accounts for uncertainty), or model-based approaches that use all available data. Deleting all cases with any missing data can waste information and bias results unless the data are truly MCAR. Replacing missing values with zeros injects artificial values that distort relationships and variance. Completing only with the available cases ignores information present in partial cases and can bias estimates unless the missingness mechanism is truly inconsequential.

Purposive sampling means choosing participants deliberately because they have specific characteristics, experiences, or knowledge that will help answer the research question. This approach focuses on information-rich cases, making it especially useful in qualitative work or when studying rare or specialized topics where random sampling would yield many irrelevant participants. Because the selection hinges on relevance rather than representativeness, the goal isn’t to generalize to the whole population but to gain deep insight from those who can speak directly to the issue. It differs from random sampling, which gives everyone an equal chance of selection; from methods that aim for equal subgroup representation like stratified or quota sampling (which seek balance across groups rather than relevance); and from convenience sampling, which uses whoever is easiest to recruit without regard to their characteristics.

Purposive sampling involves choosing participants who meet predefined inclusion criteria that are directly relevant to the research question, and doing so without randomization. This approach is nonrandom and deliberate: the researcher uses judgment to select individuals who can provide the most pertinent or rich information for studying the topic, such as experts, patients with a specific condition, or users of a particular intervention. Snowball sampling relies on referrals from initial participants, not predefined criteria-driven selection. Convenience sampling picks anyone who is readily available, which may not meet specific inclusion criteria. Quota sampling aims to fill predefined subgroups in fixed numbers, but within each subgroup the selection is typically nonrandom. The defining feature here is meeting explicit inclusion criteria through intentional, nonrandom selection, which is why this approach is the best fit.