Question 1118260: Is it problematic to change the alpha from .05 to .10?
What kind of error are we more likely to make (and why) by increasing alpha?
If we don't want to increase the alpha from .05 to .10, then what can we do to the critical regions to maintain an alpha of .05?
Answer by Boreal(15235)   (Show Source):
You can put this solution on YOUR website! It can be problematic to change alpha from 0.05 to 0.10. This means that we will increase the likelihood of rejecting Ho when in fact Ho is true. This leads to making changes in a stable system in some instances which was shown by Deming to be worse than missing a signal that there was a problem. This increases the Tyoe I error, since that is what alpha is.
I think what the answer is being looked for here is to make it a one way test, if possible, so that all the critical values are on one side of the curve, and the critical value is moved towards the mean (made smaller). This is appropriate if the direction of the change is postulated to be in one direction. It can't be done just to show significance.
Another solution to this issue is to have a bigger sample size. If the effect is real, a larger sample size will produce a larger test statistic and reach significance for the same critical value. The square root of the sample size is in the denominator of the denominator. The larger n, the smaller the denominator becomes.
|
|
|
