evaluate thesensitivity of findings to this rather strong assumption,*° a mul-tiple imputation model was fit, in which the association betweenloss to follow-up and abstinence was varied over a broad rangeof possible values (eAppendix B in Supplement 2).
INTERPRETIVE QUESTION: Missing-not-at-random is a strong assumption. If teens lost to follow-up were equally likely to have quit in both groups, this assumption understates the true effect.
he primary analysis counted eachparticipant in their originally assigned group and used a missing-not-at-random assumption in which participants lost to follow-up were coded as treatment failures (ie, vaping).
Statistical Method: STRENGTH: Conservative, intent-to-treat approach with transparent assumption.
Parameters are estimated by unweighted least squares. Time t is measured in years since the first observation in each dataset.
研究使用最小二乘法进行参数估计，时间以年为单位从每个数据集的第一个观测点开始计算。这种方法选择是统计标准做法，但未加权处理可能低估了近期数据点的重要性，因为近期数据点通常代表更先进的模型能力。时间单位的选择也影响了增长率解释的直观性。
We test for a trend over time by fitting a weighted linear model to the log-odds of usage. Under this specification, Claude is the only AI service in the survey to show a statistically significant upward trend over this period
令人惊讶的是：研究团队使用了对数几率加权线性模型来分析趋势，发现Claude是唯一一个在统计上显示出显著增长趋势的AI服务。这种复杂的统计分析方法揭示了表面上微小变化背后的真实趋势。
The difficulties of estimating long Covid | David Spiegelhalter and Anthony Masters | The Guardian. (n.d.). Retrieved September 28, 2021, from https://www.theguardian.com/theobserver/commentisfree/2021/sep/26/the-difficulties-of-estimating-long-covid
Open Science Community Tilburg on Twitter. (n.d.). Twitter. Retrieved October 28, 2020, from https://twitter.com/OpenTilburg/status/1318518990000607234
Han, H., & Dawson, K. J. (2020). JASP (Software) [Preprint]. PsyArXiv. https://doi.org/10.31234/osf.io/67dcb