Explore chapters and articles related to this topic
What are effect measures for dichotomous outcomes?
Published in Debra Evans, Making Sense of Evidence-based Practice for Nursing, 2023
When interpreting any effect measure initially, it might look like the intervention was effective, but you must consider: The role of any bias.This is one point estimate from one sample rather than a plausible range within which the true effect measure could lie with a certain degree of probability, so a confidence interval is needed to determine statistical significance or not.
Preparing Studies for Statistical Analysis
Published in Lynne M. Bianchi, Research during Medical Residency, 2022
Luke J. Rosielle, Lynne M. Bianchi
Every additional variable an investigator analyzes with inferential statistics increases the chance of a type I error (false positive). Thus, if you run enough statistical tests, eventually something is likely to turn up as statistically significant simply because some of the data happen to be a little extreme. When “fishing” for statistical significance, the probability of a type I error becomes so high it is virtually inevitable. It is much better to limit the analysis to a few pre-determined tests of the key hypothesis and report a lack of statistical significance when that is what the results indicate.
How to Develop and Write Hypotheses
Published in Lisa Chasan-Taber, Writing Grant Proposals in Epidemiology, Preventive Medicine, and Biostatistics, 2022
The terms significant and significance most commonly refer to tests of statistical significance used by most empirical studies but can also refer to clinical significance. Just as we avoid making precise statistical predictions in hypotheses (see Tip #6: Remove Any Unnecessary Words), we also avoid the use of the terms significant or significance in hypotheses. Instead, you will describe the techniques that you will use to assess statistical significance, clinical significance, as well as the role of bias and confounding in the methods section of the proposal. Statistical significance refers to the probability of observing your study's results, or results even further from the null hypothesis, if the null hypothesis was true and is typically set at p < 0.05. Statistical significance is impacted by a number of factors including the sample size of your study (i.e., the larger the sample, the more likely that the findings will be statistically significant). In addition, even if you observe statistically significant results (i.e., p < 0.05), this does not rule out that your results might be due to bias (e.g., confounding, selection bias).
Sexual satisfaction improvement in patients seeking sex therapy: evaluative study of the influence of traumas, attachment and therapeutic alliance
Published in Sexual and Relationship Therapy, 2023
Anne-Julie Lafrenaye-Dugas, Martine Hébert, Natacha Godbout
Finally, we performed analyses of variance (ANOVAs) and post-hoc analyses (Least Significant Difference method; LSD method) to test significant differences between these three groups regarding potential key correlates of improvement over the course of therapy namely the patients’ CCT, attachment representations and capacity to build therapeutic alliance. The LSD post-hoc method has the advantage of maintaining a good control over error rates when three groups are studied, even when the F-test significance is weak (Howell, 2012; Levin, Serlin, & Seaman, 1994). As the sample size brings restriction in terms of statistical power, effect sizes were used to overcome this difficulty. Indeed, some authors argue that in clinical research, practical significance, represented by effect sizes, is as important as statistical significance (e.g., Lakens, 2013). The eta2 were used to assess effect sizes, and is considered small at .01, medium at .06, and large at .14 (Fritz, Morris, & Richler, 2012). However, these comparison analyses should still be considered exploratory.
Short and long-term effectiveness of external shock wave therapy for chronic pelvic pain syndrome in men
Published in Arab Journal of Urology, 2023
Kareim Khalafalla, Ahmed Albakr, Walid El Ansari, Ahmad Majzoub, Haitham Elbardisi, Khalid AlRumaihi, Mohamed Arafa
Certainly, compared to pre-treatment levels, urinary symptoms showed intial statistically significant improvement at week 0 after treatment completion that were not sustained on the longer term (at 12 and 26 weeks). Our observed short-term ESWT effect on urinary symptoms concurs with others, where urinary symptom scores, urine flowmetry and postvoiding residual exhibited temporary early improvement after ESWT that were not statistically maintained on longer follow up [10,12,13]. We are unable to speculate the reason behind such ‘waning’ of ESWT’s effect on voiding symptoms. An important point here is the difference between statistical significance vs clinical significance. We found a significant 1.19-point improvement at week 0 compared to pre-treatment level. Nevertheless, at weeks 12 and 26 weeks the score did not regress to its baseline pre-treatment levels, but rather, still showed improvements of 0.75/0.76 points (on a 5-point scale) which despite their statistical insignificance, in our view, represent clinically significant improvements that would definitely reflect in better quality of life of patients, particularly given the number of individuals suffering CP/CPPS. Further research would benefit to undertake evaluations of the natural history of the urinary symptoms among CP/CPPS patients, as well as longer term objective measurement of ESWT’s effects on urine flow and urodynamics.
Childhood Sexual, Emotional, and Physical Abuse as Predictors of Dissociation in Adulthood
Published in Journal of Child Sexual Abuse, 2021
Mary-Anne Kate, Graham Jamieson, Warwick Middleton
ANOVA and t-tests were used to identify abuse characteristics (as measured by the BTI-R) that predicted differences between the dissociative groups and controls. For females, alpha was set to p <.001 for these analyses, and for the significant differences to be considered valid, these were required to be significantly and positively correlated to group membership which was coded on an ordinal scale (controls – 1, Uni-d – 2, Uni-DD – 3, Clin-DD – 4) to reflect the mean MID-60 score of each group being progressively higher than the last. This additional requirement prevented a situation where there were significant group differences that did not support the fundamental premise that childhood abuse is positively correlated to dissociation. Odds ratios were calculated for BTI-R items with the largest effect size (females) or t statistic (males) to examine risk factors for developing clinical levels of dissociation in adulthood. Odds ratios are used to compare the relative odds of the occurrence of the outcome of interest (i.e., clinical levels of dissociation), given exposure to the variables of interest (i.e., childhood trauma). Odds ratios determined whether a particular experience was a significant risk factor as well as the magnitude of that risk factor. In lieu of a p value, the 95% confidence interval (CI) was used as a proxy to indicate statistical significance where the lower bound of the CI does not overlap the null value, i.e., if it is greater than one (Szumilas, 2010).