What to do when you can’t find that elusive significant effect…
‘ I never failed once. It just happened to be a 2000-step process’ (Thomas Edison)
By not finding a significant effect, it doesn’t mean that you have just conducted a useless piece of research. It also doesn’t mean failure either. As the quote from Edison states above, he never failed by not finding what he was looking for. The 2000 times before he eventually “succeeded” were merely stepping stones ruling out other possible answers; this is the scientific process. So, I don’t think that purposefully manipulating your data or adding participants to gain a significant effect is within the scientific-process, however I can understand how tempting it can be, especially when finding ‘nothing’ is quite boring.
In psychology, the significant effect is sometimes the sole reason why a piece of research is conducted in the first place. Unless, it is already in a well-researched area, many people do not usually conduct a study to purposefully not find any effect at all. This may explain why after spending alot of time carrying out your research, it can become ever so easy to manipulate your data by just taking out the outliers and taking what should be in your discretion- a bit too far. One outlier becomes two, then two outliers become nine and then suddenly you have reduced your sample of 300 odd students to about 30 who conform to what you think is “doing the experiment right”. Although I think that data should be manipulated, this should only be done when it is clear that the participant has failed to follow instructions or it is clear that measurement error has occurred; this can then prevent people from purposefully manipulating their data with the sole intent of finding a significant effect.
Also, when you don’t find a significant effect, it can also be very easy to re-run the experiment with more participants. This is where it can become very easy to gain a significant effect as re-running your study with more participants can appear to others as you just replicating your study to make your results more generalisable. However, there is a difference between re-running your study and replicating your study. Re-running your study with the sole intent of finding a significant effect is unscientific and… a bit pointless. If you didn’t find a significant effect in the first place, it is highly unlikely that you will find an effect with more people. Sometimes, you are more likely to find an effect when the sample is small; This was reported in the Journal ‘Nature’ . Cambien et al (1992) found that a specific variant of a gene increased the likelihood of an individual having a heart-attack. Although the sample appeared quite large at the time (500 participants), this was nothing compared to the numerous replications that occurred after the study was published which equated to around 5000 participants; these studies did not find an effect at all. This reflects that adding more participants does not always increase your significant effect.
So what should you do if you don’t find Mr.Effect?
Nothing.
(Unless you had a ridiculously small sample of 13 people, and then you got a bit happy when it came to removing outliers, leaving you with a sample of 2)
It is perfectly reasonable, scientific and ethical to remove outliers but only with the sole intent of ridding the sample of “bad, inconsistent” data, not data that doesn’t conform to what you want it to do. You should never remove data with the sole intent of finding an effect; this is will possibly lead to a type 1 error and is also illogical, unscientific and unethical.
Instead of manipulating your data to find an effect, you should re-run your study and instead manipulate your Independent Variable in a different way; this will improve science as a whole, is a part of the scientific process and will allow science to evolve. Furthermore, it doesn’t have to be a bad thing that you haven’t found what you were looking for. In science, some of the most useful inventions and theories have been discovered by accident; Such as Penicillin, Plastic and the colour Mauve.
Also, you should never add more participants to your data with the sole intent of finding Mr. Effect, adding more participants is only logical when you have found an effect and you are replicating the study to be more generalisable. Adding participants, for the sole intent of finding an effect is a bit pointless as you are unlikely to find an effect anyway.Overall, I don’t think you should add more participants or manipulate your data to find Mr. Effect as, as I have argued above, it is unscientific.
🙂
Cambien, F., Poirier, O., Lecerf, L., Evans, A., Cambou, J., Arveiler, D., Luc, G., Bard, J., Bara, L., Ricard, S., Tiret,L., Amouyel, P., Alhenc-Gelas, F., & Soubrier, F., (1992) Deletion polymorphism in the gene for angiotensin-converting enzyme is a potent risk factor for myocardial infarction. Nature. 641-644