Causality Analysis of Climate and Ecosystem Time Series
Vyacheslav Lyubchich, Yulia R. Gel, K. Halimeda Kilbourne, Thomas J. Miller, Nathaniel K. Newlands, Adam B. Smith in Evaluating Climate Change Impacts, 2020
To study the causality of X1 → X2, the shadow manifolds, and , are reconstructed. Then, a vector on is selected, and its E + 1 nearby points on are found. The time indices of these points on are used to identify the corresponding points in (Figure 7.3). Then, a locally weighted mean is calculated to find the . The weights are found based on the distance between selected point in and its E + 1 neighbors. If the causality X1 → X2, exists, the should converge to X1(t) as value of L increases. The predictability is measured by error metrics such as mean absolute error and root mean square error (MAE and RMSE), or the correlation between observed and predicted values (ρ). This procedure is repeated over many random subsamples of the time series and the final result is the aggregation of these values over different L values (Sugihara et al, 2012).
Designing Phase II Dose-Finding Studies: Sample Size, Doses, and Dose Allocation Weights
John O’Quigley, Alexia Iasonos, Björn Bornkamp in Handbook of Methods for Designing, Monitoring, and Analyzing Dose-Finding Trials, 2017
Note that when an approach like MCP-Mod is used, the model selection should be simulated as well, as it is part of the procedure. Although this was only examined on the metric in this chapter, it should be taken into account for any other performance metric for sample size calculation. Still, similar to the case for power calculations, different sample sizes will be associated with the different true candidate models and a summary function will be needed to combine the individual n’s into a single sample size recommendation. A conservative, but often not feasible, approach is to take the maximum sample size. Alternatives would be the median or some weighted mean of the individual values.
Basic stats
O. Ajetunmobi in Making Sense of Critical Appraisal, 2021
Lower quartile = 1, that is, same value present on either side of the lower quartile. For the upper quartile, a weighted mean is taken between 5 and 6:
Daunorubicin and cytarabine for certain types of poor-prognosis acute myeloid leukemia: a systematic literature review
Published in Expert Review of Clinical Pharmacology, 2019
Juan Eduardo Megías-Vericat, David Martínez-Cuadrón, Miguel Ángel Sanz, José Luis Poveda, Pau Montesinos
From the selected studies we extracted the following data: study design, chemotherapy scheme, AML status (poor-prognosis subgroups of patients analyzed), median age, the median of CR, OS and, other survival rates, including event-free survival (EFS), disease-free survival (DFS), and relapse-free survival (RFS). For the purposes of this study, the CR rate was reported including the CR, CR with incomplete blood count recovery (CRi) and CR without platelet recovery (CRp). The median CR durations (mCRD) and the rate of allogeneic or autologous hematopoietic stem cell transplant (HSCT) were reported when data was available. The median OS (mOS) and survival rates were reported in months. In order to describe the overall effectiveness of each schedule, we calculated the weighted mean and the ranges of CR (wmCR) and OS (wmOS) of different studies adjusted by the sample size. The weighted mean was calculated by summing the product of each variable by its sample size of the different studies and dividing the result by the sum of all the sample sizes. In addition, we have represented using box-plot diagrams the CR rates and mOS reported in the different subgroups of AML with poor prognosis, as well as in the different age groups (<60 and ≥60 years) and with the dosages of DNR and Ara-C, when we have data from at least two cohorts (IBM SPSS Statistics version 22, IBM Corp., Chicago, IL, USA).
Relationship between social cognition and social behaviour following traumatic brain injury
Published in Brain Injury, 2019
In case multiple studies were identified that assessed the same social cognition domain, an average correlation was calculated based on the correlations between social cognition and post-TBI behaviour reported in the individual studies. The average correlation within a domain was calculated using meta-analytic computations. In line with recommendations by Borenstein, Hedges, Higgins and Rothstein (37), correlations (r) were first transformed to Fisher’s z scores. These scores were weighted by the inverse of the within-study variance plus the between-study variance under the random effects model. A weighted mean was calculated by dividing the sum of the weighted z scores of the individual studies by the sum of the weights. This weighted average Fisher’s z score was subsequently transformed back to a weighted average correlation (37). The sign of the individual correlations (positive or negative) was not taken into account when calculating the average correlation, only the magnitude of the correlations. The average correlation therefore represented the effect size of the association for that domain. In case multiple outcomes within the same domain were reported in the same paper (e.g. separate correlations for two measures of emotion recognition), only one correlation was included in calculation of the average correlation. In such cases, a conservative approach was taken and the lowest correlation was selected.
Disorders or Differences of Sex Development? Views of Affected Individuals on DSD Terminology
Published in The Journal of Sex Research, 2021
Elena Bennecke, Birgit Köhler, Robert Röhle, Ute Thyen, Katharina Gehrmann, Peter Lee, Anna Nordenström, Peggy Cohen-Kettenis, Clair Bouvattier, Claudia Wiesemann
A minority of the cohort chose to rate alternative terms (Table 3). Participants who rated Differences of Sex Development had a slightly better opinion about this term than participants who rated Disorder of Sex Development (Table 3). On average, (i) potential umbrella terms were rated positively (m = 5.49, pooled sd = 3.37). The weighted mean of (ii) potentially pathologizing and of (iii) non-pathologizing terms differed in 0.01 pooled standard deviations only (m = 6.24, pooled sd = 3.13 to m = 6.26, pooled sd = 3.05). (iv) Specific terms (m = 7.86, pooled sd = 2.98) were rated 0.75 pooled standard deviations better than umbrella terms.
Related Knowledge Centers
- Average
- Simpson'S Paradox
- Expected Value
- Survey Methodology
- Sampling Design
- Estimand
- Ratio Estimator
- Standard Deviation
- Standard Score
- Normalization