Identifying situations where sampling is appropriate
Any large quantitative or qualitative entity, body, aspect, issue, hypothesis, occurrence be it social, natural, economic, numeric or political can be analyzed using statistical sampling techniques. Statistical sampling entails a randomized or non-randomized selection of subsets of the whole and large population, testing the sample and making relevant and precise inferences about the entire population based on the findings obtained upon testing the sample (Westfall, 2009 p. 1).
Accounting for the wide range of deviation in the estimates of the November, 2000 Bush’s lead in the polls
Although George W. Bush held a popular lead in the polls just before the November, 2000 presidential election, during the Election Day he lost the popular vote by a small margin (nonetheless, he won the electoral vote). Prior to the November, 2000 presidential elections a wide range of estimates of Bush’s lead in the polls were given, ranging from a 12% lead to a 4% lead. This wide range of deviation in the estimates of the November, 2000 Bush’s lead in the polls can be attributed to the varied sampling techniques and tools employed by different statistical research agencies.
For instance, Kudlow gathered diverse poll results from some of these political science research groups; with CNN/USA Today Gallup poll giving 48% probable votes in favor of Bush while only 46% would select Gore, centrally to the previous weeks projection of 50 – 41 win. CBS News poll put a 43 – 41 win in favor of Bush, the Rasmussen “Portrait of America” political thermometer upheld Bush’s victory on a 43 – 31 margin over Gore (Kudlow, 2000, p. 1).
Prior political projections by statistical research agencies in the month of June did also give varied poll trends, with the LA Times crediting Bush with a 10 – point lead over Gore. The Battleground vote projection giving Bush the better part of 12 –point victory. Surprisingly though, is the diminished 4 – point lead for the republican candidate advanced by the Newsweek in the very week that followed. (Kudlow, 2000, p. 1)
Essentially this results from the fact that most of the statistical research agencies overlooked some key political-poll affective variables, especially readily availability and fair representativeness of sample correspondence. Much to his amazement, Kudlow noted that most of the polls were carried out over the weekends when political pollster turnout is dismal, narrowing the reliability of survey findings. Yet again, most statistical research agencies targeted their survey to registered voters only, and did not consider the varied political stakeholders, thus the poll projections were of little significance. It is also suspicious as to whether the statistical research surveys where rigorous enough to test a fair proportion of each state, including the proper weightings for all political stakeholders (McIntosh, 2008 p. 1).
For instance, the CNN/USA Today Gallup survey of probable presidential popularity was carried out July 14 -16, which happens to be a weekend and worse still, concentrated its survey on registered voters only. With this same practice prevailing in most of the surveys; it would be unlikely that all the research institutions would come up with comparable findings.
Sampling techniques for more accurate poll estimates
Kudlow being an economist, sought for seminal professional analysis from his close statistics practitioners in a bid to account for the wide range of deviation is the poll estimates. In which case, he received some salient empirical counsel on ensuring that correct inferences are drawn from such political projections. These diverse statistical tools and strategies employed by different statistical research agencies in the political domain account for the wide gap in the projected poll outcomes.
Key among these concerns is the sampling risk of representativeness and availability of the targeted audience, as observed by Goeas, the poll outcome and its validity is depended on, but not limited to, whether the poll was conducted during the week or over the weekend. Observing that, poll results are more reliable if the testing was done during the week.
The other strategy influencing the reliability of poll results is the selection of promising sample points from the sample space. Political surveys which target a randomized selection of all political variables are more significant than those which focus on a specific group of audience. For instance, Goeas asserts that those political polls which canvas likely voters are more reliable than those which target registered voters only (Kudlow, 2000, p, 1). A fairly randomized survey would ensure a balanced inclusion of all political stakeholders (through the use of weightings and percentages of regions) enhancing the reliability of the poll results.
However, it is important to note that, though different research agencies forwarded a wide range of poll results, one trend is common for all the survey results. It is evident from every survey that Bush holds the most popular vote and thus some reliable and accurate inferences can be drawn from weighing all the survey outcomes from different research agencies.
The Effect of a Large Sample
Since the inherent facts of the population is unknown until survey findings are analyzed and appropriate inferences drawn, the effect of a large sample on the survey outcomes and findings depends on the sampling technique employed and the distribution of the population. If a fairly randomized sampling technique is employed on an evenly distributed population, then a large sample may not necessarily affect the survey outcomes and findings. Conversely, if the population is unevenly distributed, the testing of a larger sample in the survey would ensure more accurate survey outcomes and findings.
How a Large Sample Affects the Mean
As explained in the exposition above concerning the effect of a large sample on survey outcomes, a large sample may or may not affect the mean. The point of concern is not whether the mean would increase or decrease, but the accuracy, preciseness and validity of the sample survey outcome as a representative of the whole population. In fact, depending on the nature of distribution of the population, the mean may remain steady upon considering a larger sample- when the population is evenly distributed, the mean may either increase or decrease upon testing a larger sample. A certain threshold percentage/weighting of the population must thus be tested for the sample survey to paint an accurate picture (Birchall, 2009 p. 1).
Sampling in business
In a business, sampling is an effective instrument of indentifying new and promising avenues of investment. It is also an indispensable evaluative tool for business managers in monitoring the progress of varied enterprises. It thus saves the business from unnecessary waste of its valuable time and finances. Nonetheless, some business situations such as cases of unethical behavior of employees, and the autonomy of individual differences cannot be resolved by sampling.
Reference List
Kudlow, L. (2000). The Weak-End of Polling. Web.
Westfall, L. (2009). Sampling Methods. Web.
McIntosh, J. (2008). Probability Sampling Techniques. Web.
Birchall, J. (2009). Sampling and Samples. Web.