Defaults: one or two-tailed for SEM models?

[Jennifer Gerson]

Hi everyone, 

I'm in the process of publishing an article where I used the sem command from lavaan to create a SEM model. The reviewers want me to specify whether I used a one-tailed or two-tailed test for the p-values reported in the fitmeasures summary. I've looked through the documentation but haven't had any luck finding information on the defaults for p-values. Does anyone know if lavaan sem uses one or two-tailed tests? 

If someone could point me in the right direction, I'd greatly appreciate it!

Thanks!

[Edward Rigdon]

The p values are sensible. For a statistic like the chi square, where low values are good, the p value is based on a one sided test.

Which particular p value are you concerned about?

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+unsubscribe@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[Jennifer Gerson]

Thanks for the response. The p-values the reviewer is asking about would be more along the lines of regression p-values. I had originally conducted a separate CFA and then regression models to show which personality traits predicted a certain type of social comparison. The reviewers suggested that I switch to a SEM model instead of conducting separate analyses. The p-values the reviewers are asking about are the "regression" p-values within the SEM model.

My code is below if that helps, the p-values in question are for the last 2 lines of the model2 setup.

model2<-'
Ability =~ SCOS1 + SCOS2 + SCOS3 + SCOS4 + SCOS5 + SCOS6 + SCOS11
Opinion =~ SCOS7 + SCOS8 + SCOS9 + SCOS10
Ability ~ Opinion
Ability ~ Male + AGE + AGE2 + university + RewardInterest + RewardReactivity + GoalDrive + Impulse + BISSCORE + FFFS
Opinion ~ Male + AGE + AGE2 + university + RewardInterest + RewardReactivity + GoalDrive + Impulse + BISSCORE + FFFS
'

fit2 <- sem(model2, data=mydata2)

summary(fit2, standardized = FALSE, rsq = T)

[Terrence Jorgensen]

The p-values the reviewer is asking about would be more along the lines of regression p-values.

The p values for each parameter's Wald z test is a two-sided test, assuming the null hypothesized population value = 0.  If you want to conduct a directional hypothesis test, you can check the sign of the parameter and, if it is direction of the alternative hypothesis, divide the p value by 2.

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

UvA web page: http://www.uva.nl/profile/t.d.jorgensen

[Jennifer Gerson]

Brilliant! Thank you so much!