A sample is a finite representation of a statistical group, the properties of which are used to determine facts about the whole group. We must use inferential statistics to determine the population’s characteristics by observing the sample. This technique is used when it is more practical or cheaper than a full census, but there are dangers involved in using inferential statistics.
When using a sample, the sample needs to mirror the population it comes from, but the sample may not be a precise representation. Sampling error can occur because the units selected do not accurately represent the entire population. This may occur due to chance or due to bias in selecting the sample. In either case it will result in inferences that are not accurate about the entire population. Under-coverage can happen if a group of the overall population is excluded from the sample.
No response happens when a member of the sample fails to respond, such as with a survey. Response bias can occur during interviews and surveys based upon the behavior of the interviewer or sample member or upon the wording of a survey and can cause misleading answers. Stratified random sampling can be used to help avoid sampling errors. This involves classifying the entire population into groups called strata. An equal member of each stratum is then used to identify the sample group.
Sample size is also an important consideration when inferring statistics. A sample that is too small will not accurately represent the entire population. A sample that is too large negates the benefits of sampling. The correct sample size is based upon the total population, the confidence level, and the confidence interval. The confidence interval is the margin of error, a plus or minus figure giving a range for the results. The confidence interval is how often the true percentage of the population would match your findings. In most cases the confidence level is 95 per cent.
When using a sample, the sample needs to mirror the population it comes from, but the sample may not be a precise representation. Sampling error can occur because the units selected do not accurately represent the entire population. This may occur due to chance or due to bias in selecting the sample. In either case it will result in inferences that are not accurate about the entire population. Under-coverage can happen if a group of the overall population is excluded from the sample.
No response happens when a member of the sample fails to respond, such as with a survey. Response bias can occur during interviews and surveys based upon the behavior of the interviewer or sample member or upon the wording of a survey and can cause misleading answers. Stratified random sampling can be used to help avoid sampling errors. This involves classifying the entire population into groups called strata. An equal member of each stratum is then used to identify the sample group.
Sample size is also an important consideration when inferring statistics. A sample that is too small will not accurately represent the entire population. A sample that is too large negates the benefits of sampling. The correct sample size is based upon the total population, the confidence level, and the confidence interval. The confidence interval is the margin of error, a plus or minus figure giving a range for the results. The confidence interval is how often the true percentage of the population would match your findings. In most cases the confidence level is 95 per cent.
