How to report a non-significant main effect with significant contrasts (LME)

#1
Hi :wave:,

Sorry for a lengthy first post, but I'm at my wits end :(

I am using multilevel linear modelling with nlme's lme function to analyse a mixed design (2 categorical predictors, 1 a repeated measure with 5 levels, the other not repeated with 3 levels). From Andy Field's "Discovering Statistics Using R", I read that to test the overall main effects and interactions, predictors should be added one at a time and the models compared with anova function.

I followed this train of thought, and I'm using treatment contrast, with each level compared to the base reference level. After running anova, it was shown that IV1 had a significant main effect on DV, but IV2 didn't. However, when looking at the contrasts for IV2, one (out of the 2) of the constrasts was significant, meaning, I presume, that when compared to the reference level, there was a significant difference in the DV at that level.

I understand that sometimes for ANOVA, the P value can be significant but posthoc tests fail to find the pair that are different, which may be interpreted as "there was an overall difference, but there wasn't enough power to find out where the difference was coming from"--am I correct? Now, if the overall main effect was not significant but I get a significant contrast as in my case, how can I report that? :(

Thanks in advance for any replies!

Kind regards,
xClueless

Edited to add the results here:

Code:
Model df      AIC      BIC    logLik   Test  L.Ratio p-value
baseline        1  4 1391.997 1405.308 -691.9984                        
iv1Model        2  8 1394.547 1421.170 -689.2736 1 vs 2 5.449469  0.2442
iv2Model     3 10 1392.709 1425.988 -686.3546 2 vs 3 5.837961  0.0540

Linear mixed-effects model fit by maximum likelihood
 Data: temp2 
       AIC      BIC    logLik
  1406.277 1466.179 -685.1384

Random effects:
 Formula: ~1 | subject
        (Intercept)
StdDev:      13.877

 Formula: ~1 | IV1 %in% subject
        (Intercept) Residual
StdDev:    3.658916 2.330046

Fixed effects: DV ~ IV1 + IV2 + IV1:IV2 
                           Value Std.Error  DF   t-value p-value
(Intercept)            125.84261  4.098549 146 30.704183  0.0000
IV1level2                -2.55688  1.867382 146 -1.369231  0.1730
IV1level3                -1.56102  1.847267 146 -0.845041  0.3995
IV1level4                -0.43284  1.947638 146 -0.222239  0.8244
IV1level5               -0.32706  2.041923 146 -0.160171  0.8730
IV2leve2           -5.08762  5.367947  45 -0.947778  0.3483
IV2level3            -14.88787  5.910130  45 -2.519042  0.0154
 
Last edited: