da Data interpretation is a component of modern life for most people. Interpretation is the mechanism for translating all the numerical data that we are bombarded with every minute of every day. Consumers interpret data when they turn on the television, scan headlines on an iPhone or tablet, view advertisements alleging that one product is superior to another or they make purchases based on advertising as to the price and/or efficacy of a product. A prevailing method of analyzing numerical data is known as statistical analysis and the activity associated with assessing and explaining data in order to make predictions is referred to as inferential statistics. Knowledgeable consumers understand the value of discerning the veracity of data interpretations, forecasts and recommendations by recognizing sources of bias such as sampling procedures, or misleading questions, margins of error, confidence intervals, and incomplete interpretations. The ramifications of flawed or erroneously interpreted data can be far- reaching. For example, every 10 years a major census is completed in the United States. The findings are employed to calculate the number of congressional seats that are assigned to each district; where new highways will be built; where new libraries and schools are required, where new day care centers, hospitals and nursing homes will be situated; where new parks and recreational centers will be located, and the allocation of manpower resources for fire and police departments. In the past three decades, there have been several major changes in the United States demographic. The population is aging, growing from immigration, and migrating westward and southward. The outcome has been a major shift in congressional representation. A seventeen percent drop in the Northeast coupled with a twelve percent gain in the South, the South has transitioned to a position of much greater influence in congress as a result of population based reapportionment. This is just a single example of how data gathering and the interpretation in just one arena can have enormous impact on the entire country.

In research, bias refers to inaccuracies or mistakes that crop up consistently in research analysis. By definition, bias is any propensity which hinders unprejudiced consideration of a question. In general, bias happens when systematic miscalculation is entered into sampling or testing by choosing or promoting one result or outcome over others. Bias can materialize from the method used, samples selected for the research or anything that may impact the findings positively or negatively. Bias may also manifest through publishers and organizations who furnish funding or report the research. Bias may be deliberate or unintended; if any form of bias was deliberate the researcher will typically stipulate it in the conclusions portion of his results. Sampling bias occurs when the samples of an aleatory variable are collected to ascertain its distribution are chosen incorrectly and fail to represent the true distribution as a result of non-random reasons. Ponder the following example. Assume we want to predict the winner of a presidential race by method of an opinion poll. Querying one thousand registered voters as to their expected choices would seem to offer a fairly precise prediction of the eventual winner, but only if our sample is “representative” of the electorate in its entirety . If the sample consists of only white middle class college students, then the opinions of significant portions of the electorate such as ethnic minorities, the elderly, and blue- collar workers are grossly underrepresented and our ability to accurately predict the outcome from that sample is diminished. In an unbiased sample, variations between the samples selected from a random variable and its true distribution, or variations between the samples of units from a population and the entire population they embody, should occur only from chance. If the variations are not just the result of chance, then there is a sampling bias. Sampling bias often emerges because certain values of the variable are methodically underrepresented or overrepresented in regard to the true distribution of the variable. Because of its undeviating nature, sampling bias results in a systematic slant of the approximation of the sampled probability distribution.This slant cannot be deleted simply by enlarging the size of the sample. Polling an additional thousand white college students will not enhance the predictive capacity , but querying one thousand selected at random from the electoral register would. A typical source of sampling or selection bias may be found in the blueprint of the study or in the data collection process, both of which may influence or discourage collecting data from certain groups or individual members of the groups or in certain conditions. Sampling bias is also especially pronounced whenever researchers employ sampling methods based on convenience or judgment, in which the measure used to choose samples is based on the variables of interest. In social and economic sciences, culling random samples generally necessitates a sampling frame such as an inventory of the components of the whole population, or some secondary information on some essential attributes of the target population to be sampled.
For example, administering a study about primary schools in a certain county entails acquiring a list of all the schools in the county, from which a sample can be obtained. Utilizing a sampling frame does not necessarily preclude sampling bias. There may be a failure to accurately determine the target population or outdated information might be used resulting in the exclusion of portions of the target population. Even when the sampling frame is chosen properly, sampling bias can occur from non-responsive sampling. Non-responses are quite likely to result in bias whenever the motive for non-response is correlated to the issue under scrutiny. No one understands the biased sampling that results from non-responses better than census takers and the analysts who interpret the data. In 1990, more than one third of the households that were mailed census forms did not complete and return them. When this transpires, the Census Bureau dispenses someone to the non-respondent’s home. Even with this degree of diligence, the Census Bureau was not able to connect with one out of every five of the non-respondents. Selection bias may transpire designation of the study population. The ideal study population is clearly accessible, defined, and reliable. When a study population is designated, selection biases happens when the criteria used to enlist and enroll participants into separate study panels

At first glance, this number may not seem significant. In 1990 the population was
250 million and approximately 62.5 million census forms were mailed out. Thirty five percent or 21.875 million households did not return the forms and of those , census takers could not locate and/or contact 4.375 million households. Political theorists have asserted that the majority of these were from poorer sections of large cities. This apparent biased sampling suggests that some parts of the country are over-represented in Congress and are the beneficiaries of more federal funding than they may be entitled. Sampling or selection bias often transpires in the physical and biological sciences. Bias can happen in the planning, data collection, analysis and publication phases of research. In the field of evidence-based medicine, recognizing research bias allows both physicians and laypersons to analyze and assess scientific literature and eschew treatments which may be ineffective or possibly harmful. Bias can happen at any stage of research including study method or data collection as well as in the phase of data analysis and publication. Bias is not a binary variable. Analysis of bias cannot be limited to a single query: is bias evident or not?
Alternatively, readers of the research literature must take into account the degree to which bias was minimized by precise design and implementation. As some degree of bias is almost always the case in a published study, readers must also judge how bias might impact a study’s findings. Origin of pre-trial bias includes inaccuracies in study design and in participant recruitment. These inaccuracies can result in fatal flaws in the data which cannot be corrected during data analysis. The delineation of risk and outcome should be clearly stipulated prior to study implementation. Subjective measures can have significant inter-rater deviation and the arbitrary limits may make differentiating between groups challenging. This can

overstate the observed variance making a statistically meaningful result less probable.
Unbiased, validated risk stratification models or standardized outcome measures should have reduced inter-rater deviation and are more suitable for use. Data collection methods may include imaging data, interviews, laboratory data, medical chart review, physical exams, and questionnaires. Standardized procedures for data collection, including education for study personnel, can reduce inter-observer variability particularly when multiple staff persons are gathering and inputting the data. Blinding of study personnel to the participant’s history and outcome status can also reduce bias. Similarly, blinded examiners can also scrutinize imaging data and validate diagnoses without examining participants. Selection bias may transpire during determination of the study population. The optimal study population is accessible, clearly defined and reliable. When a study population is established, selection bias happens when the criteria employed to enlist and enroll participants into separate study panels is