Significance level / I didn't make my own thread

[jmochogi]

I am doing a pearson product moment correlation but the results come with alpha=0.01. Can I change it to 0.05?

[trinker]

Alpha is what you set. P value is what the computer calculates as the significance level. You (the researcher) compares the pvalue to your alpha. If's less than your alpha the results are significant.

[Ventures]

Just a couple of quick points.

You can have any Alpha level you like. In reality a large majority of people use 0.05 through tradition and custom. There is no statistical reason that I can see that we use 0.05 rather than 0.01 or 0.1

A result may be statistically significant but a little bit of common sense can go a long way in determining the practical significance of a result.

[Dason]

I moved the discussion to its own thread since it wasn't related to the thread you posted in.

[trinker]

There is no statistical reason that I can see that we use 0.05 rather than 0.01 or 0.1

I would disagree slightly with this point. It is true that .05 is convention but the alpha level is the researcher's accepted threshold for making a type I error. This should be set with each experiment a priori, and really depends on numerous things, such as previous research in the area, use of results etc. but is sometimes influenced by a researcher's need to be published a posteriori. So, I'd say in a hamburger taste test a researcher chosen .1 may be fine but for a new cancer drug we'd want to see .05 or .01 alpha level. Sometimes alpha is called the confidence level. This is perhaps a better name for alpha as it better describes what it is.