Sampling and the non statistician.

1. Introduction
2. Sampling
3. Factors that cause variations
4. Factors that polarize opinions
5. Sampling goals
6. Method
7. Questionnaires

Introduction
Sampling plays a very important role in almost every aspect of human endeavor from medicine, politics, agriculture, material sciences insurance and actuarial sciences to, of course, business management, where it is extensively employed in production planning, quality control, auditing, financial management and market research, to name a few. When opinions are sampled, this is known as “opinion sampling” or simply “polling”. Throughout this article,"opinions" and "survey results" are used interchangeably.

In many instances, we want to find out people's reaction to say, a new product, an impending election, an issue to which a lot of money is to be committed, the market reception of a company's impending IPO. or an outcome which should prove valuable if known.
In some other cases, information and statistical data is needed by the insurance industry. Knowledge of this usually spells the difference between success and failure.

Experience in an industry may be an informal source of knowledge, which some rely upon, or with which they dispute some results of statistical data sampling.

There is published Government statistics, from where information can be obtained. This can be employed to supplement a sampling exercise, where knowledge of market size is desired.
This may however constitute a challenge in some third world countries where some of these published information may not be accurate. Other sources of information which can complement such published data are from:-

• Insurance companies,
• Telecoms companies for subscriber data and habits
• Retail store chains for Selling or buying patterns.
• Old fashioned post/zip codes in the Postal systems.
• Email/Internet service providers (linking such to zip/post codes for effective email marketing).

Successful sampling therefore will reduce a need to carry out detailed sampling on the whole population. For example, the market size of a product can be estimated, if the population of the area is known through published demographics, and the proportion of users in a given sample area is extrapolated for the whole population.

This has hitherto, been the exclusive specialty of Statisticians. However, since the arrival of the PC revolution, the availability of specialized statistical packages and spreadsheet programs have actually reduced the arcane processes of analyzing several types of data to form meaningful information into just a few strokes of the keyboard. This does not still override the need for a knowledge of statistical methods.

Some factors may cause variations in sampling results;

1. Varying experiences among the constituent communities: Preference for a type of broadband service for example will vary with the topography of their areas. Communities with high rise buildings or mountains will obviously show preference for wired services as opposed hot spots, where obstructions by the mountains will reduce effectiveness.
2. Variations in their goals and aspirations. For example, employers of labor will not definitely see labor policies the same way employees see such.

1. Average incomes. An example may be the issue of brand name watches. Wealthier people would like exclusive or customized watches while the less wealthy would be attracted by imitations. Low income people would rather like something functional.
2. Living standards. This would influence, for example, the type and models of automobiles desired. Those with higher incomes would prefer luxury cars in preference to functional ones demanded by those with lower income levels. A market research for a type of car may show some variations in those areas.
3. Education levels. People will expect to have jobs and appointments that utilize their educational qualifications. Their assessment of a government or consumer confidence will vary with those who's jobs do not require much depth in education.
4. Profession or industry of interest. Specialized articles, journals or magazines will only be of interest to some in a particular profession.
5. Religious beliefs: These may cause variations in perception and usefulness of products.

A prudent pollster will therefore use these as classifications when carrying out a sampling or polling exercise.
Trying to survey a market for solar power panels or wind power systems for those living in apartments or condos in a crowded city center for example, may not yield good results as some of those apartments may not have an easy access to direct sunlight for the solar panels, or space for the wind power systems. Email marketing may therefore produce many negative responses as it is not associated with a type of living accommodation.
A pollster should therefore focus efforts towards those who live in town houses or have access to adequate space, by way of Zip/post codes.

1. Locating a sampling area that typifies the whole population. In US poll surveys for example, Peoria, Illinois is usually regarded as the average American city. Surveys or voting patterns there is estimated to represent the typical pattern nationwide.
2. Extreme. One can look for a location whereby the results show the worst case, in which your sampling is expected to beat.
3. Boundary data. If it happens here, then it must be. If it fails here, then it must fail.
   This may be two separate group of samples obtained from two different locations which may represent the most optimistic and pessimistic scenarios respectively.

Method
Verification of data.

1. Trial sampling with known data. We may therefore test data samples with what already exists in these data sources.
2. Sub sampling. We may also take samples among samples which have produced known results in the mentioned sources. For example, sampling an internet market among users of handsets of a particular provider.
3. When collecting data, there must be a reference.
   You want to launch a new product, and would like to know how it stacks up against an existing product, or at what price it will be accepted in the existing location, a poll is needed. The existing product acts as reference.

You can them compare such with the statistics of the regulator, with dates in mind as the results may be dynamic.
This gives some level of confidence towards getting good results with the least margin of error with your sampling.
The inclusion of a date in the field enables double information.
Whenever the date indicated in the public data precedes those of the sampled data this will form the basis for another survey on market growth, which in effect, can be carried out simultaneously with the market research.

Many polling exercises fail as a result of over reliance on questionnaires. Most busy executives cannot afford to spare the time to read them carefully and answer the questions, nor ready to volunteer information.
On line incentives, like a software utility or program do not necessarily help as some, out of eagerness to receive the incentives just give answers without really thinking much about it.

The pollster approaches the busy executive to introduce the product through a sample or a brochure.