So in my class last night we tried to reproduce the statistical tests from a 2014 psych science paper with an open data badge. The OSF page had the raw data but no codebook or script/syntax, but it wasn't too hard to figure out. The original authors did a 2 x 2 x 2 x 2 ANOVA. (1)

There were two "marginally significant" main effects (p = .07ish) and they reported that the predicted interaction between two vars was significant with p = .034. When we tried to reproduce the F ratio sig test we got p = .068. (2)

Is there a chance they just / the p by 2 like with a one-tailed t test? What else could account for this? We ran it in SPSS and I'm not sure how they ran it -- the data came in an .xls file (3)