simple power analysis

iiro

New Member
#1
I am doing my master thesis about publication bias and QRP. One option would be to test individual papers with test for excess success (TES) and then see how the TES scores correlate with some factor (e.g. importance of the findings) that could predict QRP and Publication bias.

Now I am trying to calculate a priori power. Below I tried to calculate sample size for 0.75 power, if theres r=.15 effect size. Please let me know if you know whether its correct.

t tests - Linear multiple regression: Fixed model, single regression coefficient

Analysis:A priori: Compute required sample size
Input:Tail(s) =One
Effect size f² =0,6129032
α err prob =0,05
Power (1-β err prob) =0,75
Number of predictors =2
Output:Noncentrality parameter δ =2,5965237
Critical t =1,8595480
Df =8
Total sample size =11
Actual power =0,7653745
 

hlsmith

Omega Contributor
#2
What is QRP?

Also, there is literature showing that conducting post hoc power calculations has many issues. I am sorry I don't have a citation readily available for this statement. Are you trying to conduct post hoc power calculations?
 

iiro

New Member
#3
What is QRP?

Also, there is literature showing that conducting post hoc power calculations has many issues. I am sorry I don't have a citation readily available for this statement. Are you trying to conduct post hoc power calculations?
Thanks for your response! QRP = questionable research practises.Yes, TES is a technique to check how probable some findings are if the study was not affected by publication bias or QRPs.

Yes I have read some debates whether the TES technique makes sense, but I am not an expert in statistics thus its hard to have an educated opinion.
 

hlsmith

Omega Contributor
#4
Another amusing issue that has been repeatedly illustrated is the distribution of reported p-values from published articles and how there is an unnatural increased density in p-values just below 0.05.