prpnw

http://fr4nw.wordpress.com/2011/10/14/ethical-principles/#comments

http://psucd3.wordpress.com/2011/10/27/is-it-dishonest-to-remove-outliers-from-the-data/#comments

http://prpsjj.wordpress.com/2011/10/26/the-pursuit-of-ethical-conduct-can-often-hinder-the-development-of-knowledge/#comments

http://roydeanschlipp.wordpress.com/2011/10/28/is-validity-important-in-your-research/#comments

Thanks 😀

Grubbs (1969) defined an outlier as “An outlying observation, or outlier, is one that appears to deviate markedly from other members of the sample in which it occurs.” So for example if your data looked like this… 12,12,13,14,16, 32 it is obvious to see that 32 is the outlier as it deviates far from the other numbers that are all relatively close together. However, outliers are not always as easy to spot as the one in the example and it can take a while to find them in large amounts of data.

Outliers can occur due to participant or measurement error. Quite often it could just be caused by chance. Take for example a simple reaction time test; outliers could occur for a number of reasons. Some participants may have been unwilling to co-operate and may have chosen to disregard the instructions they had been given. Or perhaps a participant was unclear on the study and what they had been asked to do and so did not perform the way the researcher had hoped. Measurement error could have also occurred through equipment not working correctly or precisely.

One of the main problems with outliers is that it can increase or decrease the mean, obscuring the interpretation of your results. If your data contains outliers then the generalisability of your findings will decrease.

So, if outliers can affect your findings should you extract them from your data or is this dishonest? In my opinion it depends on the circumstances as to whether you should. Say for example the equipment wasn’t working correctly in a particular block, I think it would be acceptable to extract the outliers in this situation because otherwise the validity of the experiment would decrease. It would no longer be measuring what it was meant to be and therefore the findings would be invalid. Zimmerman (1994) suggested that it is wrong to keep outliers as it inflates error and gives inaccurate meanings to the data.

On the other hand, I do not believe that outliers should be deleted when they are real pieces of data that just don’t fit in with the ‘norm’. For example, if you were carrying out a study on feet size and someone has extremely small feet it shouldn’t be treated as an error and you shouldn’t delete this data just because it deviates from the mean. If you remove people from your data because of reasons like this then your findings will no longer fit the population; it won’t be a true representation.

In conclusion, I think it should be down to your own discretion as to whether you should remove outliers. However if you do remove them I think you should give a valid reason why in your research report.

Reliability and validity are both used for assessing the quality of a measurement procedure. They are linked in the way that a procedure cannot be valid unless it is reliable. Because of this some would argue that reliability is more important the validity. But is this true? I beg to differ.  

The reliability of a measurement procedure is how well it produces constant results over time. If a procedure is thought to be reliable it means that it would produce near to identical results every time the exact same experiment was repeated on the same individual. An example is the Stanford-Binet IQ test…if you took this over and over you would get near enough the same result every time.

An experimenter doesn’t want their measurement procedure to differ each time it is used as it would contain a large amount of error and obviously hinder the results. For example, picture Javelin in the Olympics…each person’s throw is measured; if the equipment used for measuring the distance thrown isn’t reliable it means the measurement will hold a considerable amount of error.

Roger. L. Martin suggested that reliability and validity seem to conflict… “more reliability requires fewer variables and therefore less validity, and vice versa”.

The validity of a measurement procedure is how much the experiment measures what it is supposed to. Say for example a researcher decides that they want to measure the pulse rate of participants who are exposed to a horror movie. Is the experiment measuring an increase in pulse rate from the horror movie or an increase from the nerves of being in an experimental environment? This is an example of the simplest definition of validity; face validity. A lack of validity will make it much harder for results to be generalised and this would be a massive flaw when concluding your results. However sometimes high validity can be a disadvantage as participants can sometimes guess what is being tested and change their answers to what they think the experimenter wants to hear (Webb, Campbell, Schwartz, Sechrest, & Grove, 1981).

Although some people see reliability and validity as having the same importance others may disagree. If a study has low face validity you can easily state this in the research report and then create a new hypothesis from what you actually found and develop further on this. However if a study has low reliability, the whole study will contain errors and the end result will be misleading.

Take this hair product advertisement http://www.youtube.com/watch?v=qSgOlz7ZYr0. How reliable and valid do you think it is? The product claims to target weak, limp, lifeless, dull and straw like hair and also states that over 73% of 356 agree. However, how valid is this statement? The data was actually collected through survey and therefore self report, so how were limp, lifeless etc measured? Surely by using self report it wasn’t measured reliably which also makes you question whether the results are valid. Did the product actually improve all those flaws or did it just make the hair feel softer for the first few days? As you can see from highlighting parts of this advertisement both reliability and validity are equally important as otherwise they hinder the results of the experiment.

The example above just demonstrates how significant reliability and validity are. In my opinion neither of them are more important than the other as they are both necessary for an experiment to be successful and worthwhile. Without validity and reliability the study would hold much more error and the results couldn’t be generalised.

Statistics helps form explanations by the collection and analysis of data. It consists of both basic and complex analysis depending on what is being interpreted.

If you have a hypothesis that you want to prove or disprove then statistics is the answer. Once you have run your experiment and collected your data you will want to interpret it, right? It’s no good having a bunch of meaningless numbers, you need to take it a step further and analyse your results. If you have a statistical background you will be able to take your analysis of the data further than if you had little knowledge of the subject. It will be easier to see what patterns you are looking for, what statistical analysis to use (a T-test, Anova etc), and what you can interpret from the results. 

Admittedly even without statistical knowledge you can spot some patterns in the data you have collected, you can see from looking at graphs that there is a correlation between the purchase of ice creams and the weather for example. It wouldn’t take Einstein to work out that more people buy ice cream when the weather is hot. Data can be simple unless you choose to complicate it.

 However, you will be limited to how much you can explain from what you have collected.  If you don’t have background knowledge of statistics you wouldn’t be able to produce the P value or know which statistical procedure to use. This could hinder your findings and prohibit you from exploring them further.

In conclusion, yes, statistics is needed to explore your data if you want to prove or disprove a hypothesis. Without statistical procedures there is only so far you can go with the exploration and interpretation of your results. Statistics makes data more meaningful and easier to understand so that it can be used to further scientific research.

Statistics are used in a range of fields not just the typical mathematics, psychology and other academic subjects. Unsurprisingly statistics are all around us which is why having a statistical background is valuable and versatile.

Statistics is significant to psychological research; it allows you to explore and analyse research using tools such as an Anova and T-tests. Without these tools you would not be able to reject or accept a hypothesis or even find significance in your data.  Statistics is a useful and specific way of presenting information which could normally take paragraphs to explain. 

Even advertising companies have jumped on the band wagon and use statistics to sell their products or services. For someone with a strong statistical background it may be a lot easier to find flaws in statements made about certain products. It’s all good and well to want to buy a product that 90% of those surveyed would recommend, until you read the small print and realise that the sample size consists of 10 people. Those who have the advantage of being familiar with stats would instantly look for this ‘hidden information attempt’ in order to determine the reliability of the statistic given.  Steven Wright cleverly stated that ‘47.3% of all statistics are made up on the spot.’

Although having a statistical background has many advantages (no matter how tedious stats may seem) they are some disadvantages too. Psychologists or keen mathematicians for example may become too obsessed with trying to find a significance in their data that they completely miss the bigger picture and concentrate less on what they initially set out to find. Statistics is also complex and can be a lengthy tool to use.  

In my opinion, although statistics may sometimes be tedious and yawn provoking it is becoming more and more desirable (I think so) to obtain background knowledge of the tool. It is useful in many academic subjects as well as everyday life and allows you to ‘look outside the box’.