{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreidn5qs5lvi57imfm5pwddnqsq4scv2vpguaoc6gaxjc5d2qobaotu",
    "uri": "at://did:plc:wwyqal4cnqhuwyacdj7rqq3n/app.bsky.feed.post/3mioq6givank2"
  },
  "path": "/t/is-there-a-contradiction-between-the-results-of-the-mixed-effects-model-and-those-of-the-pairwise-comparisons/28698#post_1",
  "publishedAt": "2026-04-04T14:35:46.000Z",
  "site": "https://discourse.datamethods.org",
  "textContent": "I am comparing two groups across six different time points using a mixed-effects model. No statistically significant group-by-time interaction, which is my primary interest. Here is the code:\n\n\n    model1 = lmer(score ~ group * time + BMI + gender + prothesis.design + PCL + (1 | patient), data = data.frame, REML = T, contrasts = list(group = contr.sum, time = contr.sum))\n    Anova(model1, type = 3, test.statistic = \"F\")\n\n\nHowever, when I perform pairwise comparisons at each time point for data presentation and graphing, the first time point shows a significant difference (the adjusted confidence interval does not include zero). Here is the code:\n\n\n    emm.hss = emmeans(model1, ~ group * time, adjust = \"bonferroni\")\n    pairs(emm.hss, by = \"time\", adjust = \"bonferroni\")\n    confint(pairs(emm.hss, by = \"time\", adjust = \"bonferroni\"))\n\n\nIs this a contradiction or a comparison based on different metrics? If it is not a contradiction, which result should I prioritize in my reporting, and how should I address it?\n\nThanks in advance.",
  "title": "Is there a contradiction between the results of the mixed-effects model and those of the pairwise comparisons?"
}