An outlying observation, or outlier, is one that appears to deviate markedly from other members of the sample in which it occurs (Grubbs, 1969). Outliers may appear in Psychology researches, and affecting the analysis of the data into conclusions. This decreases the validity of the results, making it hard to generalise and apply to the target areas.

 To encounter the problem, researchers would normally eliminate the extreme scores. This will make the total set of data to have a clearer sense of direction on the effects of relationships. However, excluding outliers is a reductionist approach, and may prevent us to develop our understanding by removing a special case of data. Furthermore, it is possible that the researchers may not include outliers as to increase the validity of the research. This is researchers biased.

In some cases, outliers can appear because of errors. The errors can be created from many parts of the research. Outliers can be created by few individual participants not understanding the procedures, and provide extreme scores when responding to the conditions. This can be solved by conducting a pilot study before the research. A pilot study is a rehearsal of the research in a small scale, and it helps to identify any possible problems the research may face when being carried out in a larger scale.

Outliers can also be created when there are errors in analysing and concluding the data. An example can be seen with Quantitative Data. It is possible that the set of data has been put in incorrectly during calculation, and affect the results of Mean, Median and Standard Deviation. A good method to deal with the problem is to reject all values beyond a criterion number of standard deviations from the mean of each experimental cell; typically discarding all values beyond about 2 to 3 SD of the mean (Selst & Jolicoeur,1994). Researchers may also use Inter-quartile range of the data set to remove extreme scores, by only focusing on the mid 25-75%. It is useful because according to the Normal Distribution, majority of the population lie in the middle range, with the minority of population lie within the extreme ends.

Outliers can be problems to researches, by being the unconventional data which differ from the majority. It is difficult to decide if it is appropriate to remove them. The advantage of removing outliers is that it makes it easier to evaluate the data, without extreme scores. The disadvantage of removing them is that it may make us ignore some special but important data, stopping us to provoke with developing an improved idea on the targeted topic.

Reference:

Selst, M.V. & Jolicoeur,P. (1994): A solution to the effect of sample size on outlier elimination, The Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology, 47:3,632.

Grubbs, F. E. (1969): Procedures for detecting outlying observations in samples. Technometrics 11, 1–21.

Advertisements

About eugenekitfung

An undergraduate student in Bangor University who lives in Normal Site. NORMAL SITE, TRA LA LA LA!

5 responses »

  1. Rory says:

    Although all your points about outliers are valid, with research to back up your points, there is not a clear line of argument that you are following in terms of your own views on outliers.
    I would argue outliers should be omitted from results dependent on the size of the sample being tested. If, for instance, there was one outlier in a sample size of 10, this should be considered significant and should be included in the results. However, if the sample size was a lot larger, and there was a small amount of outliers, these should then be omitted as they are far from the average scores.
    I agree to when you say pilot studies should be used, and a good briefing would also ensure any outliers are more likely to be due to the fact the participant was particularly unique.

  2. ln1992 says:

    Pilot studies are definitely a perfect way to test for any errors within the experiment that may produce outliers or effect all of the data collected. The outlier debate is a difficult area of psychology, and I must admit I sit on the fence in terms if whether they should be removed or left. However I agree with Rory in that it all depends on sample size. I think it also depends on the extent to which the outliers are being removed, for example if they are being removed just to shape the results to what the experimenter expected or if it is purely because it could be participant error. If outliers exist there should be a method to go through to find out why the outliers exist and this should determine whether or not they should be removed.

  3. I definitely agree with Rory that whether or not outliers should be removed from the results of an experiment should all be due to sample size. If the outlier is thought to be a genuine error on the participants behalf than I believe that it should be acceptable to remove. However it is a different case when an experimenter removes an outlier to manipulate the results of an experiment to his/her advantage/idea of what the results should look and should not be a case of when it is acceptable to remove outliers, simply because no experiment where the results have been changed to meet an experimenters expectations represents the true results of the study. It an outlier is taken out I believe that there should be research done into why these outliers exist and it should then be determined by the results whether or not it would be suitable to remove them.

  4. Josh Webb says:

    An intresting blog and you highlight areas of debatable intrest within the field. I particularly like how the reductionistic disadvantage of removing outliers is highlighted. Perhaps there could have been more debate on the issue of outliers, and whether it is important to note their existance. The minds of each individual are each different and perhaps a result should not be immediately discredited due to it not following a specific pattern. The use of pilot studies is an effective idea to remove outliers due to the test, I agree.

    (feel free to comment on my blog – http://psychrmblog.blogspot.com)

  5. Zak says:

    Whether an outlier is be omitted or not should depend on the context of the research. As stated above, using a large sample means an outlier can affect the validity of the results – and thus be a hindrance to research. However, sometimes researchers are hoping to find these outliers providing that it is not due to error. An example of this would be testing people for disorders. If 100 random people were given the Sally Anne test, and only one of those 100 was a person with autism – they would appear as an outlier. However, in this context, this outlier would be very useful to the researcher.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s