EXAMPLE ANSWERS: Q21

Q21. Investigating related material sources on the internet, discuss how bias can be avoided in scientific research.

Learning outcomes: 5.3

Student answer As discussed in q19, in order for a research to be scientific it has to fulfilL certain criteria which include a design where there is ability to measure response before and after an intervention; participants should be randomly assigned to control or experimental groups and a hypothesis developed to measure the effect on the dependent variable by manipulating an independent variable However to be 'scientific' it is important to ensure that all variables have been considered. One of the most important to eliminate or control- as best as possible- is the element of research or experimenter bias. This can take several formats; but in short it is important to ensure that the researcher is not influencing the outcome- consciously or unconsciously- in either the design or the reporting. It is fair to say that it can be difficult to completely eliminate bias, but by recognising and accounting for it, the effects can be minimised. There are three main areas where bias can occur: the design or methodology, the data collection and the reporting Design: in order to minimise design bias the principles of scientific experiment design should be followed: 1. Control; by including a variable which can be kept constant we ensure that it affects the response in the same way for all aspects of the experiment. This in turn reduces experimental error which is an outcome which is due to causes other than the aspect being investigated. This could take the format of a control group exposed to the same conditions other than the experimental condition e.g. 3 sets of people watching the same film one in manipulated conditions and one in severe heat, one in severe cold. 2. Replication; this relates to more than one person or group being assigned to an experimental condition. It means than any results cannot be due to fluke and that the results can be generalised to a wider population. 3. Randomization- assigning people or events to a random group/setting. This reduces or eliminates selection or sampling bias where participants or areas for study are selected or excluded depending on opportunity or on perspective. This could be interviewing passers by in a venue with no public transport links (so are more likely to live locally or have a car) about a bus route or conducting research via home telephone numbers during the working day which may self-select a sample. In order to address this, a varied sample should be selected – using a random sampling strategy or multiple days/times. Where this is not possible, a rationale should be provided e.g. this is a random selection of responses from an opportunist sampling in Tipton during a weekday afternoon. Where a wider sampling strategy is used there is more opportunity to extend results as a generalisable result to the wider population. 4. Blocking; attempting to balance out external factors e.g. matching people as similar and then one member of each pair is randomly assigned to a treatment condition and the other is randomly assigned to a control group. Where the experiment designer does not recognise that bias does or can exist, they are unable to account for it or for all possible variables. Data collection Bias within data collection can take place in two main ways; inaccuracy or procedural. Inaccuracy can include failure to correctly calibrate a tool. This can be a measurement instrument such as a weighing scale or a qualitative scale such as rating from 1-5. The very wording on the scale "which did you most prefer...indicating that you did prefer, can have a powerful influence and wording used should be examined for the unconscious bias. In the same way the scale should be kept consistent - always using the poles in the same way. If during statistical analysis there is zero error it may be a sign that the researcher needs to re-examine the tool used. In the same way the procedure used for research should be consistent- what is done, said or shown to one group must be done shown or said to the other. Spoken word which is not corded can be subject to bias from the subtle perceptions of the researcher. This may be as simple as the unconscious prejudice or beliefs of the researcher often displayed via body language or tone or may include coercion or placing pressure on subjects to behave in a certain way e.g. asking parents to respond quickly to research which needs to show that attending a playgroup has a positive impact on emotional intelligence or the playgroup will be closed. It can be very hard to design a hypothesis that you wish to test without there being some interest or presupposition of how the outcome will be. The use of a null hypothesis and the cross checking of a readers design by an ethics committee can often help reduce bias in this area. A further example of procedural error is when subjects attempt to please or second guess the researcher or are reluctant to Espinosa (express?) in a way they may feel will be judged. For example many people will not give an honest answer if they feel they will appear racist. Hearing people use "I'm not x but..." in a response may indicate that there is bias in the system. Using a combination of methods such as interview and questionnaire or may also assist to reduce bias. Reporting bias: this can take two main forms, partiality and publication. When writing results it is important to thoroughly examine all of the data, the methods and the findings. If there is bias this must be acknowledged e.g. this is a random selection of responses from an opportunist sampling in Tipton during a weekday afternoon. It is extremely hard to keep an open mind but this is the requirement of the scientific method- not allowing the data to be skewed but to 'speak for itself'. In this way all data must be included, outliers should be investigated and reported or recorded as appropriate. Maintaining partiality can be difficult and there is a reliance on the reporter to be objective and to accurately report any evidence – supporting or not. This can be difficult depending on the funding stream (e.g. many PhDs are funded by pharmaceutical companies and research that does not support the anticipated view are often ‘lost’ or published in such a way that the general population will never see the information. Overall there are many ways that bias can influence research but careful consideration at each of the stages reduces the possibility or impact and allows for a confidence to be given to scientific or academic research. (LO 5.3 An understanding of 'bias' in research)

Tutor feedback (DO NOT delete/edit feedback. Write amendments, additional information & thoughts underneath this table) Correct. A highly comprehensive answer to this one which tackles the question of bias and its impact and gives clear understanding of what bias actually is, thank you. I very much applaud not only your appropriate answer, though also your examples throughout.

(Further student work goes here if requested through feedback)

Further Student Example Answer:  The internet is full of material covering every imaginable subject, the vast majority of this information is unchecked or verified. Ranging from phishing and hacking sites through aggressive sales sites to the prestigious scientific community sites. Sales websites often pose as official looking research sites in order to give an air of authority to their work whilst glossing over the fact their research is, at best, bias towards their product.

Wikipedia is often the first reference site used by the public when researching a subject, however Wikipedia is an open source site, which allows anyone to post information about a subject, and allows others to verify or report inaccuracies. The vast majority of the information on Wikipedia is quite accurate however it should not be trusted implicitly and further checks on its content must be undertaken before accepting its information such as cross referencing with other sites and sources.

Once you have identified a reliable source of information you should look to the following to give an indication as to the accuracy of the content.

The methodology of any research experiments must be open and transparent, with a clearly defined purpose, they must also be carried out in a scientific way in that they must be controlled conditions with repeatable and clearly defined results which are measurable utilizing scientific standards, the test size should also be statically large enough to ensure a true representation of the effects (7 people out of 10 selected volunteers doesn’t not truly reflect the efficacy of a medicine or treatment etc. designed to be given to millions of people) Check their case studies for validity and accurate recording of results making sure the evidence is scientific and not anecdotal.

The testing process must also include the use where possible of double blind testing as well as the use of a control set to be in place to ensure a true and accurate result to exclude the potential for human bias on the part of the subjects or the researchers tainting the results. Ensure the research has been validated and reviewed by other experts in the field, checking on any referencing given.

These methods are fairly standard in almost all research facilities, however when surfing the internet for information it is always best to obtain information from several sources where possible and also to verify each source independently to ensure the quality of their information.

Q21. PLEASE RESUBMIT. More information required. You have provided a clear understanding of inaccuracies of information on the internet and warnings as to anomalies encountered and also the idea of commercial bias, where a salesperson may sway information towards a commercial end, however please provide detail of a broader view of scientific bias.

Below are most of the accepted Research Bias’s known today, some are deliberate others accidental and some unavoidable.

Research bias

Is where the experiments are set up in a way that can influence the results in order to produce a certain desired outcome.

Some bias in research arises from experiment errors and either fails to take into account or even discounts all of the unexpected or unwanted variables.

Design Bias

Design bias arises when researchers select subjects that are more likely to generate the desired results, a reversal of the normal processes governing science.

Also design bias can be introduced when the researcher fails to take into account the inherent biases liable in most types of experiment. Some bias is inevitable, and the researcher must show that they understood this and tried their best to lessen the impact, or take it into account in the statistics and analysis.

Reporting Bias

Is a form of bias that occurs after the research is finished and the results analysed.

For example, research into the health benefits of Acai berries may neglect the researcher’s awareness of limitations in the sample group. The group tested may have been deliberately selected as being all female, or all over a certain age to provide favourable results.

To counter the above biases the true intent of the project must be acknowledged and accepted or factored into resulting reports.

Measurement Bias

Error in the data collection and/or the process of measuring can occur resulting in Measurement bias misrepresenting the actual results of the experiment.

Instrument bias is one of the most common sources of measurement bias in quantitative experiments. This is the reason why instruments should be properly calibrated before and after each test, and multiple samples taken to eliminate any obviously flawed or aberrant results.

In a quantitative experiment, a faulty scale would cause an instrument bias and invalidate the entire experiment.

In qualitative research, the scope for bias is wider and much more subtle, requiring the researcher to be constantly aware of the problems.

Psychological Bias

People are often extremely reluctant to give socially unacceptable answers, for fear of being judged. For example, someone may strive to avoid appearing homophobic or racist in an interview.

This can skew the results, and is one reason why researchers often use a combination of interviews, with an anonymous questionnaire, in order to minimize measurement bias.

Studies of behaviour are difficult, as the act of participating in the studies, performing the research will actually have an effect upon the behaviour of the sample groups.

This is unavoidable, and the researcher must attempt to assess the potential effect.

Procedural Bias

Procedural bias is where an unfair amount of pressure is applied to the subjects, forcing them to complete their responses quickly.

i.e. employees asked to fill out a questionnaire during their break period are more likely to rush, rather than reading the questions properly.

Using students forced to volunteer prior to being released after class is another type of research bias, and they are more than likely to fill the survey in quickly, leaving plenty of time for other activities.

Interviewer Bias

This is one of the most difficult research biases to avoid in many quantitative experiments when relying upon interviews. The interviewer may subconsciously give subtle clues in with body language, or tone of voice, that subtly influence the subject into giving answers skewed towards the interviewer’s own opinions, prejudices and values.

Response Bias

Conversely, response bias is where the subject consciously, or subconsciously, gives response that they think that the interviewer wants to hear. The subject may also believe that they understand the experiment and are aware of the expected findings, so adapt their responses to suit.

Again, this type of bias must be factored into the experiment, or the amount of information given to the subject must be restricted, to prevent them from understanding the full extent of the research

Any experimental design must factor the above Psychological Bias’s into account, or use some form of anonymous process to eliminate the worst effects.

Quantitative Research Bias:

Generally thought to be free of any Bias due to large number of experiments producing a wide selection of results producing a higher quantity of research data, however any and all of the experiments can be bias in the ways listed above.

Qualitative Research Bias:

A smaller number of experiments run to a higher standard of data collection processes thought to be more open to interpretation. For example, when using social research subjects, it is far easier to become attached to a certain viewpoint, jeopardizing impartiality.

The main point to remember with bias is that, in many disciplines and processes, it is unavoidable.

Selection/Sampling Bias

Sampling bias occurs when the process of sampling actually introduces an inherent bias into the study. There are two types of sampling bias, based around those samples that you omit, and those that you include:

Omission Bias

This research bias occurs when certain groups are omitted from the sample. An example might be that ethnic minorities are excluded or, conversely, only ethnic minorities are studied.

For example, a study into heart disease that used only white males, generally volunteers, cannot be extrapolated to the entire population, which includes women and other ethnic groups.

Omission bias is often unavoidable, so the researchers have to incorporate and account for this bias in the experimental design.

Inclusive Bias

Inclusive bias occurs when samples are selected for convenience.

This type of bias is often a result of convenience where, for example, volunteers are the only group available, and they tend to fit a narrow demographic range.

There is no problem with it, as long as the researchers are aware that they cannot extrapolate the results to fit the entire population.

Reporting Bias

Reporting Bias is where an error is made in the way that the results are disseminated in the literature. With the growth of the internet, this type of bias is becoming a greater source of concern.

The main source of this type of bias arises because positive research tends to be reported much more often than research where the null hypothesis is upheld. Increasingly, research companies bury some research, trying to publicise favourable findings.

Unfortunately, for many types of studies, such as meta-analysis, the negative results are just as important to the statistics.

Acceptance and Acknowledgment of Bias is essential in any and all experiments as well as conducting open intent well throughout research experiments utilising Double Blind experiment procedures, extensive data recording and record keeping should assist in producing impartial quality results

Correct, many thanks for this, it is a very thorough report on the various biases that exist in such experimentation and research and shows your comprehensive understanding of bias.