Basic Research Design:
Lynne M. Bianchi in Research during Medical Residency, 2022
Statistical analysis: Data are typically analyzed using inferential statistics that allow the investigator to formally test a hypothesis and make comparisons between groups. The statistical tests used will be determined by the types of measurements made (categorical, continuous) and the number of groups in the study. For example, when the data from each group have a normal distribution and the variables measured are continuous (interval, ratio) then parametric tests, such as Student'st test, Pearson's correlation coefficient, or analysis of variance (ANOVA), might be used. When the data are not normally distributed or the variables are categorical (nominal, ordinal), then an appropriate non-parametric test will be used, such as the Chi-squared test, Fisher's exact test, Spearman's rank correlation coefficient, or Mann-WhitneyUtest.
Non Parametric Methods
Marcello Pagano, Kimberlee Gauvreau in Principles of Biostatistics, 2018
Nonparametric methods also have a number of disadvantages. If the assumptions underlying a parametric test are satisfied, the nonparametric test is less powerful than the comparable parametric technique. This means that if the null hypothesis is false, the nonparametric test would require a larger sample to provide sufficient evidence to re‐ ject it. If the data do come from an underlying normal population, the power of the Wilcoxon tests is approximately 95% of that of the t‐tests; if the t‐test requires 19 ob‐ servations to achieve a particular power, the Wilcoxon test requires 20 observations to achieve the same power. In addition, the hypotheses tested by nonparametric techniques tend to be less specific than those tested by traditional methods. Because they rely on ranks rather than on the actual values of the observations, nonparametric tests do not use everything that is known about a distribution. This, of course, presumes that our in‐ formation about the underlying population is correct. Finally, if a large proportion of the observations are tied, σT and σW overestimate the standard deviations of T and W, respectively. To compensate for this, a correction term must be added to the calculations [3].
Nonparametric Methods
Marcello Pagano, Kimberlee Gauvreau, Heather Mattie in Principles of Biostatistics, 2022
Nonparametric methods also have disadvantages. If the assumptions underlying a parametric test are satisfied, the nonparametric test is less powerful than the comparable parametric technique. This means that if the null hypothesis is false, the nonparametric test would require a larger sample to provide sufficient evidence to reject it. This loss of power is not substantial, however. If the sample data do come from an underlying normal population, the power of the Wilcoxon tests is approximately 95% of that for the t tests. In other words, if the t test requires 19 observations to achieve a particular level of power, the Wilcoxon test would need 20 observations to have the same power. Another disadvantage is that the hypotheses tested by nonparametric techniques tend to be less specific than those tested by traditional methods, focusing on medians rather than means. Because they rely on ranks rather than on the actual values of the observations, nonparametric tests do not use all the information known about a distribution. This, of course, presumes that our information about the underlying population is correct. Finally, if a large proportion of the observations are tied, then σT and σW overestimate the standard deviations of T and W, respectively. To compensate for this, a correction term must be added to the calculations [222]. These correction terms are beyond the scope of this text, but they are implemented in most statistical packages that perform nonparametric hypothesis tests.
Evaluation of the relationship between para-aortic adipose tissue and ascending aortic diameter using a new method
Published in Acta Cardiologica, 2022
Adem Adar, Orhan Onalan, Fahri Cakan, Hakan Keles, Ertan Akbay, Sinan Akıncı, Ali Coner, Cevahir Haberal, Haldun Muderrisoglu
Data were analysed using the SPSS 23 (IBM Statistical Package for Social Sciences version 23) software package. Categorical variables were expressed using frequency distributions and numerical variables using descriptive statistics (mean ± standard deviation). The Kolmogorov–Smirnov test was used to determine whether the data conformed to the normal distribution. Subsequently, a parametric test was used to analyse normally distributed data and data that did not confirm to the normal distribution were analysed using a nonparametric test. The independent samples t-test and Mann–Whitney U test were used to check whether there was a difference between the measurements of the two independent groups. Additionally, multivariate logistic regression analyses were conducted to assess the relationship between PAT and AAD. In multivariate regression models, the effect size was adjusted for variables with a significance level ≤ 0.10 in the univariate analysis. Adjusted odds ratios (ORs) and their corresponding confidence intervals (CI) were given. 2-tailed probability (p) values of< 0.05 were considered statistically significant.ICC was used to determine the in-class reliability of PAT measurement (95% CI).
Unintended imitation of fundamental frequency in female speakers
Published in Logopedics Phoniatrics Vocology, 2020
Laura Lund Kuhlmann, Jenny Iwarsson
Despite the low number of subjects included in the study, we decided to statistically analyze the dataset in order to see whether any tendencies were strong enough to still reach statistical significance. As a principle, parametric statistics are based on the assumption that the sample is drawn from a normally distributed population. A risk when analyzing such a small dataset with parametric statistics is the type 2 error, meaning that the statistical analysis falsely infers the absence of an effect that is. The analysis was done using vassarstats.net [21]. By means of a two-factor analysis of variance (ANOVA) with repeated measures on both factors, data were analyzed with regard to (1) the difference in Fo between conditions (baseline, repetition), (2) the differences in Fo between word type (words, nonwords), and (3) the interaction effect between condition and word type.
Patrolling the Boundaries of Gender: Beliefs, Attitudes and Behaviors Toward Trans and Gender Diverse People in Portuguese Adolescents
Published in International Journal of Sexual Health, 2020
María Victoria Carrera-Fernández, Ana Almeida, Xosé Manuel Cid-Fernández, Pablo Vallejo-Medina, Yolanda Rodríguez-Castro
EQS 6.1 software was used to evaluate the confirmatory factor analysis (CFA). The maximum likelihood method estimation with robust standard errors (ML,R) was used since multivariate normality was not met. The overall fit indices used were the Root Mean Square Error Approximation (RMSEA) with a 90% confidence interval and the Comparative Fit Index (CFI). Values below 0.06 in the RMSEA and greater than 0.95 in the CFI were considered indicative of a good fit. We followed the recommendations of Jackson, Gillaspy, and Purc-Stephenson (2009). Other results were obtained with SPSS software. The omega coefficient and its confidence interval were obtained using R (R Core Team, 2017) and the psych package (Revelle). Total scores met all the assumptions for a parametric test and no outliers were observed. A t-test was the parametric test chosen for comparisons.
Related Knowledge Centers
- NONparametric Statistics
- Standard Deviation
- Aggregated Distribution