0

I use lmerTest to obtain p values for my lmer models in R. Unfortunately, running the model multiple times give me different p values each time. Is there a way running the model multiple times to calculate one robust p value? Any other way to obtain more robust p values are also highly welcome.

My code (this is a model with best random structure according Aicc, interactions with high varioation inflation factor were removed):

model <- lmer(cmic~h2of+pH+TempGlob+exp.age+tree.density+log.div+
              h2of:pH+
              h2of:TempGlob+
              h2of:log.div+
              pH:log.div+
              TempGlob:log.div+
              exp.age:log.div+
              (1|experiment/site/block),data=micc2,REML=TRUE)

Thanks S

Schweine Mareike
  • 51
  • 3
  • 1
    Please show some code. Normally the Satterthwaite approximation is the default and that should always yield the same p-value. You seem to have managed to use parametric bootstrap. If that is the case, the answer might be increasing the number of bootstrap resamples or getting more data. – Roland Jul 26 '18 at 12:35
  • How can I increase the number of bootstrap resamples? The dataset includes 1000 data points in total. I think this should be enough?! – Schweine Mareike Jul 26 '18 at 13:00
  • How are you calling `summary`? As I said, it should be deterministic by default. – Roland Jul 26 '18 at 13:02
  • Actually, I use anova(model). Summary(model) is the better wax to go than, I guess. – Schweine Mareike Jul 26 '18 at 13:15
  • No, `anova` should also be deterministic by default. You still have not shown what you actually do. – Roland Jul 26 '18 at 14:01

0 Answers0