Statistical programming, matrix languages, and more

Using simulation to compute p-values

Accepted Solution Solved
Reply
Contributor
Posts: 39
Accepted Solution

Using simulation to compute p-values

Dear all, Dear Rick,

I simulated 20,000 data of OLS coefficients under the null that the coefficient is equal to zero.

Now I have 20000 simulated data and one observed coefficient.

I want to calculate the p - value.( The probability of getting the results I did  given that the null hypothesis is true)

In Rick Wicklin's book, he suggested for chi-square procedure:

pval=sum(Q>=qObs)/Numsample;

But in my case is it right? Isn't it p-value for one-tailed test ?

Do I need to take

pval=2*sum(Q>=qObs)/Numsample;

Thank's,

Orit


Accepted Solutions
Solution
‎07-14-2014 08:44 AM
SAS Super FREQ
Posts: 3,381

Re: Using simulation to compute p-values

It sounds like you want a two-sided test. For a discussion, see How to compute p-values for a bootstrap distribution - The DO Loop .  For a two-sided example (and code), see p. 14 of this paper: http://support.sas.com/resources/papers/proceedings10/329-2010.pdf

View solution in original post


All Replies
Solution
‎07-14-2014 08:44 AM
SAS Super FREQ
Posts: 3,381

Re: Using simulation to compute p-values

It sounds like you want a two-sided test. For a discussion, see How to compute p-values for a bootstrap distribution - The DO Loop .  For a two-sided example (and code), see p. 14 of this paper: http://support.sas.com/resources/papers/proceedings10/329-2010.pdf

☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 1 reply
  • 252 views
  • 0 likes
  • 2 in conversation