Logistic Regression. A Self-Learning Text. Third Edition ed. ed. New York: Springer; 2010.
Logistic Regression. Data were analysed using R. Continuous demographic characteristics were compared by Student’s t- Test or Xxxx-Xxxxxxx U test. Categorical variables were compared by chi-squared or Xxxxxx’x exact test. The primary outcome, whether smoking increases risk of ALS, was analysed using logistic regression with maximum likelihood estimation. We generated 8 models with combinations of one categorical and one continuous measure of smoking, comparing the Akaike Information Criterion (AIC) of the models to assess fit (Akaike, 1998). Odds ratios were adjusted for age, educational attainment, gender and alcohol consumption. Assuming an odds ratio of 1.8, a 20% smoking rate in the control population and alpha of 0.05, we had 71% power with a sample size of 400 cases and controls in a 1:1 ratio.
Logistic Regression. For index/usage pairings that chose the best-ARI partition in more than 40% of cases, we evaluated the influence of clustering algorithm used to create partitions and data design factors on correct choice by the index/usage. Logistic regression models were fit using generalized estimating equations to estimate variance to account for the repeated measures study design. Models included two-way interactions of data design factors with clustering algorithm. Number of underlying subtypes. The Xxxxxx-Xxxxxxx (SAD) index performed significantly better for data with six underlying subtypes. All other indices performed worse for data sets with six underlying subtypes. Distribution across subtypes. Most indexes performed slightly better for data with one subtype larger than the others (distribution level 2). Distribution level 2 had a significant negative effect on the Xxxxxx- Xxxxxxx (SAD) and Xxxxx-Xxxxxx indices.
Logistic Regression. Logistic regression was performed (Table 4a) between discrimination and the outcome of perceived stress, using ‘High Discrimination’ as the reference, which demonstrated a statistically significant relationship (p < . 046) meaning experiencing lower levels of discrimination is associated with lowered odds in reporting higher levels of perceived stress. Multiple-variable logistic regression was performed utilizing perceived stress as the outcome and significant variables from Xxxxxx’x Exact test (employment status, birthplace, age 30-49 and 50+). None were found to be significantly associated. However, interactions did not demonstrate any significant impact on the OR of the discrimination variable (OR, 0.24; CI: 0.048, 1.24). An interesting relationship to not despite lack of significance is that the 30-49 age group and 50+ carry an increased risk of outcomes of perceived stress, (OR, 1.58; CI: 0.23, 11.02); (OR, 1.1; CI: 0.14, 8.26) while being born outside of the US, (OR, 0.61, CI: 0.09, 4.11) decreases risk of perceived stress outcomes. Logistic regression for acculturative stress and perceived stress were not significant (Table 4c) (p < .43), however, an OR of 0.6 suggests lowered odds of experiencing perceived stress as a result of lower levels of acculturative stress. Multiple-variable regression with sociodemographic variables and acculturative stress were also not reported to have a significant relationship. *3 missing
Logistic Regression. Table 3.4 presents the results of the logistic regression model identifying predictors of couple-level supportiveness for condom use. Men‟s desire for a child resulted in greater likelihood of being in a relationship with low supportiveness (OR: 1.3; CI: 1.1-1.4; p<0.01). Men and women who perceived that the HIV negative partner was at high risk for infection in the next 12 months had an increased likelihood for low couple-level supportiveness for condom use. Men‟s perceived risk resulted in a 3.0 fold increase in odds of being in a relationship with low supportiveness (OR: 3.0; CI: 1.8-5.1; p<0.001) while women‟s perceived risk resulted in 2.2 times increase (OR: 2.2; CI: 1.4- 3.6; p<0.01). Men and women who believed that it is acceptable for a husband to force his wife to have unprotected sex were far more likely to have low couple-level supportiveness, with even greater risk when reported by the woman (OR: 3.9; CI: 1.1-8.8; p<0.01) in comparison to the man (OR: 2.6; CI: 1.1-6.3.; p<0.05). Women who reported condom breakage in the last three months were also more likely to be in a relationship with low supportiveness (OR: 1.7; CI: 1.1-2.7; p<0.05) as were women who believed there was a drug that an HIV positive person could take to prevent transmission (OR: 2.5; CI: 1.6-5.4; p<0.05). Finally, for couples in Zambia, the odds of being in a relationship with low supportiveness towards condom use was 1.6 compared to Rwandan couples (OR: 1.6; CI: 1.1-2.4; p<0.05). .
Logistic Regression. Table Three displays the survey adjusted results from the logistic regression. The outcome variable was whether or not a child or adolescent was obese. The first model has only hours of sleep as a predictor of whether or not a child is obese. For each additional hour of sleep the odds of obesity were 15.8% lower. The final model controls for age, sex, food insufficiency, the income level for the family, physical activity, and screen time. After controlling for all of these factors, there is a statistically significant relationship between hours of sleep and the odds of obesity. For each additional hour of sleep, there were 15.2% lower odds of obesity among children and adolescents aged ten to seventeen. For example, a child who sleeps 9 hours compared to a child who sleeps 8 hours has 15.2% lower odds of obesity after controlling for the factors above. This indicates that children who have more hours of sleep have lower odds of obesity and that sleep could potentially be a place for intervention. It is unknown whether or not children have lower odds of obesity because they sleep more, or if they sleep more because they are not obese, and further research to establish which comes first is needed. This, however, is a good starting place for establishing the link between sleep and obesity in children and adolescents in the United States. Of interest, whether or not the child goes to bed at the same time produced varied results for the outcome variable. When compared to those who always went to bed at the same time, children and adolescents who usually went to bed around the same time on weeknights had 24.7% lower odds of obesity. For those who only sometimes had the same bedtime during the week, the odds of obesity were 16.4% lower than those who always had the same bedtime. Lastly, for those who rarely or never went to bed at the same time on weeknights, the odds of obesity were 16.8% lower than those who always did. Results with respect to confounding variables such as physical activity and screen time were as expected. Children who got less physical activity had higher odds of obesity than those who had more physical activity, and those who spent more time watching TV or on electronic devices also had higher odds of obesity than those who had less screen time. Table Three: Odds of Obesity from Logistic Regression. Multinomial Regression Table Four displays the relative risk of remaining underweight, overweight, or obese for each factor when compared to a c...
Logistic Regression. The adjusted logistic regression models explained 49% of the variance (Adjusted R2 =0.49) of the intention to test soil and 50% of the variance (Adjusted R2=0.50) of intention to wash hands (Table 4). TPB variables accounted for 34% and 37% of soil testing and handwashing intention respectively. Variables that were significantly associated with intention to test soil and wash hands, via chi-square analysis, were garden context, income, age, race, education, garden region, gardener chemical practice method and garden site history. These variables were therefore adjusted for in the logistic regression model. All TPB variables were statistically significant in the soil testing and handwashing logistic regression models. The odds of soil testing intention increased with a positive attitude (aOR=4.13, 95% CI: 2.31, 7.36), stronger subjective norms (aOR=3.02 95% CI: 1.82, 5.01), and higher PBC (aOR=1.86 , 95% CI: 1.09, 3.15). The odds of handwashing intention increased by three-fold with an increase in positive attitudes (aOR= 3.09, 95% CI: 1.30, 7.33) and perceived behavioral control (aOR=2.92 95% CI: 1.23, 6.96). Subjective norms had strongest influence on handwashing intention (aOR=4.58, 95% CI= 2.27, 9.25). Education and garden chemical practices also had a statistically significant influence on both behaviors. Gardeners with less than a college education were significantly more likely to have a higher intention to test their soil (OR=0.11, 95% CI: 0.05, 0.25) and wash their hands (OR=0.23, 95% CI: 0.08, 0.61). The odds of soil testing and handwashing intention decrease a factor of 0.20 and 0.10, respectively for gardeners that use pesticides compared to gardeners that use natural (without use of pesticides but not USDA certified organic) methods, holding other variables constant.
Logistic Regression. The naive Bayes training algorithm relies on conditional independence for the features used. This is however not the case for the features described in this model; of the four features capturing relative polarity for example, one of them is true and the rest is false on any given entry of the data set. A paper by Ng and Jordan (2002) argues that when a training set approaches infinity, logistic regression will consistently outperform naive Bayes in performance. Though the AMI corpus is far from infinite, it is useful to research the use of this training algorithm and if it can outperform the Bernoulli naive Bayes in this case. Furthermore, moving away from Bernoulli naive Bayes allows for the use of more dynamic features that do not have to be only true or false. This enables the use of features that contain scales, and the use of PCA, which relies on normalised data with a mean of 0 to work. 4 Data preprocessing Running logistic regression on the model without any preprocessing results in drastically lower performance compared to Bernoulli naive Bayes, resulting in an F1 score of 48.93 compared to 60.83. However, when looking at earlier work on this topic by X. Xxxx, Yaman, Precoda, Richey, and Xxxxxxx (2011), downsampling or upsamling the data to adjust data imbalance can have a significant impact on performance. In the case of the AMI corpus, data imbalance is significant as the proportion of acceptances to rejections is over 10:1. In the next section, both removing acceptances to downsample and duplicating rejections to upsample are documented.
