PSYC1001 Lecture Notes - Lecture 6: Statistical Significance, Null Hypothesis, Statistical Power
![](https://new-preview-html.oneclass.com/aOyR3GWKrXY5QpV8p057jldMxEP7v8o0/bg1.png)
Practical significance and statistical power
Significance when < 0.05 or 5%?
• Critical cut off convention/decision rule
• Rejects hyp when <0.05, result is due to chance
• Implies made an error 5% of the time
Types of error
Screen clipping taken: 7/06/2017 3:56 PM
Type 1 error/alpha
• Probability of rejecting null hypothesis if true
• Error we have been deciding on
Type 2 error/beta
• Probability of retaining null hypothesis when false
• Error that inevitability blows out
• Cannot set nor accurately determine
• As one reduces the other one grows
Statistical power = 1-beta
• Probability of correctly rejecting a null hypothesis which is false
• Depends on same factors as beta
find more resources at oneclass.com
find more resources at oneclass.com
![](https://new-preview-html.oneclass.com/aOyR3GWKrXY5QpV8p057jldMxEP7v8o0/bg2.png)
Normal decision
rule (p<0.05)
Conservative
decision rule
(p<0.0001)
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Significance when < 0. 05 or 5%: critical cut off convention/decision rule, rejects hyp when <0. 05, result is due to chance. Implies made an error 5% of the time. Type 1 error/alpha probability of rejecting null hypothesis if true: error we have been deciding on. Type 2 error/beta: probability of retaining null hypothesis when false, error that inevitability blows out, cannot set nor accurately determine, as one reduces the other one grows. Statistical power = 1-beta: probability of correctly rejecting a null hypothesis which is false, depends on same factors as beta. Factors affecting power (1-b: variability of effect you"re looking for (more variability = lower power) Size of effect looking for (larger effect = more power) Sample size (larger sample size = more power) If too low = too little power, not sensitive enough to detect an efect of importance. If too many = too much power, study too sensitive, tiny effects of no practical significance might be found.