Academia.eduAcademia.edu

Figure 3 – uploaded by Josh Reiss

See full PDFdownloadDownload figure
Table 3. Statistical analysis of data from [63]. Statistically significant results at « = 0.05 are given in bold.

Table 3 Statistical analysis of data from [63]. Statistically significant results at « = 0.05 are given in bold.

Related Figures (8)

Table 1. Summary and classification of high resolution audio listening tests.
Table 1. Summary and classification of high resolution audio listening tests.
Table 2. A. List of studies included in meta-analysis. B. Risks of potential biases and other issues in the included studies (see Sec. 2.5). Low risk “-”; unclear risk “2”; high risk “[2].” The last column identifies if these risks tend strongly towards false positives (Type I errors), false negatives (Type II errors) or neither (Neutral). C. Total number of trials and correct answers for each study, with the associated binomial probability (see Sec. 3.1). may contribute towards Type II errors, i.e., an inability to demonstrate discrimination of high resolution audio. If no one was able to distinguish between formats and there were no issues in the experimental design, then all trial results would be independent, regardless of whether the tri- als were by the same participant, and regardless of how par- ticipants are categorized. But [63] also gave a breakdown of correct answers by gender, age, audio experience, and hearing ability, depicted in Table 3. Non-audiophiles, in par- ticular, have very low success rates, 30 out of 87, which has a probability of only (p(X< = 30) = 0.25%). Chi squared analysis comparing audiophiles with non-audiophiles gives ap value of 0.18%, suggesting that it is extremely unlikely that the data for these two groups are independent. Simi- larly, analysis suggests that the results for those with and
Table 2. A. List of studies included in meta-analysis. B. Risks of potential biases and other issues in the included studies (see Sec. 2.5). Low risk “-”; unclear risk “2”; high risk “[2].” The last column identifies if these risks tend strongly towards false positives (Type I errors), false negatives (Type II errors) or neither (Neutral). C. Total number of trials and correct answers for each study, with the associated binomial probability (see Sec. 3.1). may contribute towards Type II errors, i.e., an inability to demonstrate discrimination of high resolution audio. If no one was able to distinguish between formats and there were no issues in the experimental design, then all trial results would be independent, regardless of whether the tri- als were by the same participant, and regardless of how par- ticipants are categorized. But [63] also gave a breakdown of correct answers by gender, age, audio experience, and hearing ability, depicted in Table 3. Non-audiophiles, in par- ticular, have very low success rates, 30 out of 87, which has a probability of only (p(X< = 30) = 0.25%). Chi squared analysis comparing audiophiles with non-audiophiles gives ap value of 0.18%, suggesting that it is extremely unlikely that the data for these two groups are independent. Simi- larly, analysis suggests that the results for those with and
Fig. 2. A forest plot of studies where mean and standard deviation over all participants can be obtained, divided into subgroups based on whether participants were trained. negatives). We identified an unclear risk in each category if the risk had not been addressed or discussed and a high risk if there was strong evidence of a flaw or bias in a category. Potential biases led both to Type I and Type II errors, i.e., to falsely suggesting an ability to discriminate or not to discriminate high resolution content, though Type II errors were more common. Furthermore, biases often existed that might result in Type I errors even when the overall result demonstrated an effect (e.g., [59]). The most common way that results are presented in the studies are as the mean percentage of trials with correct discrimination of stimuli averaged over all participants. Thus this effect measure, equivalent to a mean difference [77], is used in most of the analysis that follows. The influence of these and other choices will be analyzed in Sec. 3.7.
Fig. 2. A forest plot of studies where mean and standard deviation over all participants can be obtained, divided into subgroups based on whether participants were trained. negatives). We identified an unclear risk in each category if the risk had not been addressed or discussed and a high risk if there was strong evidence of a flaw or bias in a category. Potential biases led both to Type I and Type II errors, i.e., to falsely suggesting an ability to discriminate or not to discriminate high resolution content, though Type II errors were more common. Furthermore, biases often existed that might result in Type I errors even when the overall result demonstrated an effect (e.g., [59]). The most common way that results are presented in the studies are as the mean percentage of trials with correct discrimination of stimuli averaged over all participants. Thus this effect measure, equivalent to a mean difference [77], is used in most of the analysis that follows. The influence of these and other choices will be analyzed in Sec. 3.7.
Table 4. Multiple comparisons testing. The last two rows tested the probability of a lack of significant results in multiple comparisons. The last column considers the probability of obtaining at least (or at most, for Nishiguchi 2005 and Meyer 2007) tha many significant results given the significance level and number of tests. 3 META-ANALYSIS RESULTS
Table 4. Multiple comparisons testing. The last two rows tested the probability of a lack of significant results in multiple comparisons. The last column considers the probability of obtaining at least (or at most, for Nishiguchi 2005 and Meyer 2007) tha many significant results given the significance level and number of tests. 3 META-ANALYSIS RESULTS
Fig. 3. Funnel plots of the 17 studies where a mean difference per participant was obtained, along with associated 95% confidence intervals. Much of the asymmetry in the overall funnel plot that might be attributed to publication bias is removed when funnel plots are given for subgroups of studies with and without training.
Fig. 3. Funnel plots of the 17 studies where a mean difference per participant was obtained, along with associated 95% confidence intervals. Much of the asymmetry in the overall funnel plot that might be attributed to publication bias is removed when funnel plots are given for subgroups of studies with and without training.
Table 5. Sensitivity analysis showing the percent correct (effect estimate) and confidence intervals under different approaches to the meta-analysis. Data type was considered as either continuous (CONT) for means and standard errors over all participants, or dichotomous (DIC) for number of correct responses out of all trials. Inverse Variance (IV) and Mantel-Haenszel (MH) statistical methods were considered.
Table 5. Sensitivity analysis showing the percent correct (effect estimate) and confidence intervals under different approaches to the meta-analysis. Data type was considered as either continuous (CONT) for means and standard errors over all participants, or dichotomous (DIC) for number of correct responses out of all trials. Inverse Variance (IV) and Mantel-Haenszel (MH) statistical methods were considered.
Dr. Reiss has published over 160 scientific papers, in- cluding more than 70 AES publications. His co-authored publication, “Loudness Measurement of Multitrack Audio Content Using Modifications of ITU-R BS.1770,” was re- cipient of the 134th AES Convention’s Best Peer-Reviewed Paper Award. He co-authored the textbook Audio Effects: Theory, Implementation and Application. He is co-founder of the start-up company LandR, providing intelligent tools for audio production. He is a former governor of the AES, and was Chair of the 128th, Papers Chair of the 130th, and Co-Papers Chair of the 138th AES Conventions. Josh Reiss is a Reader with Queen Mary University of London’s Centre for Digital Music, where he leads the audio engineering research team. He received his Ph.D. from Georgia Tech, specializing in analysis of nonlinear systems. His early work on sigma delta modulators led to patents and an IEEE best paper award nomination. He has investigated music retrieval systems, time scaling and pitch shifting techniques, polyphonic music transcription, oudspeaker design, automatic mixing, sound synthesis, and digital audio effects. His primary focus of research, which ties together many of the above topics, is on the use of state-of-the-art signal processing techniques for professional sound engineering.
Dr. Reiss has published over 160 scientific papers, in- cluding more than 70 AES publications. His co-authored publication, “Loudness Measurement of Multitrack Audio Content Using Modifications of ITU-R BS.1770,” was re- cipient of the 134th AES Convention’s Best Peer-Reviewed Paper Award. He co-authored the textbook Audio Effects: Theory, Implementation and Application. He is co-founder of the start-up company LandR, providing intelligent tools for audio production. He is a former governor of the AES, and was Chair of the 128th, Papers Chair of the 130th, and Co-Papers Chair of the 138th AES Conventions. Josh Reiss is a Reader with Queen Mary University of London’s Centre for Digital Music, where he leads the audio engineering research team. He received his Ph.D. from Georgia Tech, specializing in analysis of nonlinear systems. His early work on sigma delta modulators led to patents and an IEEE best paper award nomination. He has investigated music retrieval systems, time scaling and pitch shifting techniques, polyphonic music transcription, oudspeaker design, automatic mixing, sound synthesis, and digital audio effects. His primary focus of research, which ties together many of the above topics, is on the use of state-of-the-art signal processing techniques for professional sound engineering.
Related topics:
Audio EngineeringMeta-Analysis and Systematic Review
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved