Limits...
A Permutation Test for Unbalanced Paired Comparisons of Global Field Power.

Files BT, Lawhern VJ, Ries AJ, Marathe AR - Brain Topogr (2016)

Bottom Line: Global field power is a valuable summary of multi-channel electroencephalography data.The results show that the proposed test finds global field power differences in the classical P3 range; the other tests find differences in that range but also at other times including at times before stimulus onset.These results are interpreted as showing that the proposed test is valid and sensitive to real within-subject differences in global field power in multi-subject unbalanced data.

View Article: PubMed Central - PubMed

Affiliation: U.S. Army Research Laboratory, Aberdeen Proving Ground, MD, USA. benjamin.t.files.civ@mail.mil.

ABSTRACT
Global field power is a valuable summary of multi-channel electroencephalography data. However, global field power is biased by the noise typical of electroencephalography experiments, so comparisons of global field power on data with unequal noise are invalid. Here, we demonstrate the relationship between the number of trials that contribute to a global field power measure and the expected value of that global field power measure. We also introduce a statistical testing procedure that can be used for multi-subject, repeated-measures (also called within-subjects) comparisons of global field power when the number of trials per condition is unequal across conditions. Simulations demonstrate the effect of unequal trial numbers on global field power comparisons and show the validity of the proposed test in contrast to conventional approaches. Finally, the proposed test and two alternative tests are applied to data collected in a rapid serial visual presentation target detection experiment. The results show that the proposed test finds global field power differences in the classical P3 range; the other tests find differences in that range but also at other times including at times before stimulus onset. These results are interpreted as showing that the proposed test is valid and sensitive to real within-subject differences in global field power in multi-subject unbalanced data.

No MeSH data available.


Related in: MedlinePlus

Results of three statistical tests applied to experimental data. In all panels, the dark blue line depicts the measured group mean GFP difference of Condition A, target images, minus Condition B, background images. The three tests are the unbalanced paired permutation test, conventional paired T test and the conventional paired permutation test. For the two permutation tests, the shaded area covers the central 95 % of the permutation distribution. Values falling outside of that area correspond to a rejection of the  hypothesis at the uncorrected .05 criterion. For the T test, the shaded area depicts a 95 % confidence interval about the measured value. Samples at which the confidence area does not include 0 correspond to a rejection of the  hypothesis at the uncorrected .05 criterion. For all tests, pale marks at y = 0 indicate samples at which the  hypothesis is rejected at the uncorrected .05 criterion, and dark marks indicate samples at which the  hypothesis is rejected after FDR correction. All tests were done with N = 13 subjects
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4829629&req=5

Fig4: Results of three statistical tests applied to experimental data. In all panels, the dark blue line depicts the measured group mean GFP difference of Condition A, target images, minus Condition B, background images. The three tests are the unbalanced paired permutation test, conventional paired T test and the conventional paired permutation test. For the two permutation tests, the shaded area covers the central 95 % of the permutation distribution. Values falling outside of that area correspond to a rejection of the hypothesis at the uncorrected .05 criterion. For the T test, the shaded area depicts a 95 % confidence interval about the measured value. Samples at which the confidence area does not include 0 correspond to a rejection of the hypothesis at the uncorrected .05 criterion. For all tests, pale marks at y = 0 indicate samples at which the hypothesis is rejected at the uncorrected .05 criterion, and dark marks indicate samples at which the hypothesis is rejected after FDR correction. All tests were done with N = 13 subjects

Mentions: This experiment used EEG data from 13 subjects performing a target detection task in a rapid serial visual presentation paradigm. After artifact rejection, an average of 48.5 (range 25–54) target trials and 426.1 (range 240–496) non-target trials were maintained. In the terms used in Experiments 1 and 2, data imbalance as the proportion of target trials was 1/9.8 on average (range 1/8.8–1/10.6). The three statistical tests (unbalanced paired permutation, conventional paired-samples permutation, paired-samples T test) were applied to the GFP at each time point from 1000 ms before stimulus onset to 2000 ms after onset. At the sampling rate of 256 Hz, this results in 768 time points. Results of the four statistical tests are shown in Fig. 4. All three tests rejected their respective hypotheses (q < .05, FDR corrected) from 296 to 1051 ms post-stimulus onset. However, consistent with the simulation results from Experiment 1, the conventional statistical tests rejected their respective hypotheses prior to stimulus onset 239 and 235 times out of 256 total pre-stimulus samples for paired-samples T test and paired-samples permutation test, respectively.Fig. 4


A Permutation Test for Unbalanced Paired Comparisons of Global Field Power.

Files BT, Lawhern VJ, Ries AJ, Marathe AR - Brain Topogr (2016)

Results of three statistical tests applied to experimental data. In all panels, the dark blue line depicts the measured group mean GFP difference of Condition A, target images, minus Condition B, background images. The three tests are the unbalanced paired permutation test, conventional paired T test and the conventional paired permutation test. For the two permutation tests, the shaded area covers the central 95 % of the permutation distribution. Values falling outside of that area correspond to a rejection of the  hypothesis at the uncorrected .05 criterion. For the T test, the shaded area depicts a 95 % confidence interval about the measured value. Samples at which the confidence area does not include 0 correspond to a rejection of the  hypothesis at the uncorrected .05 criterion. For all tests, pale marks at y = 0 indicate samples at which the  hypothesis is rejected at the uncorrected .05 criterion, and dark marks indicate samples at which the  hypothesis is rejected after FDR correction. All tests were done with N = 13 subjects
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4829629&req=5

Fig4: Results of three statistical tests applied to experimental data. In all panels, the dark blue line depicts the measured group mean GFP difference of Condition A, target images, minus Condition B, background images. The three tests are the unbalanced paired permutation test, conventional paired T test and the conventional paired permutation test. For the two permutation tests, the shaded area covers the central 95 % of the permutation distribution. Values falling outside of that area correspond to a rejection of the hypothesis at the uncorrected .05 criterion. For the T test, the shaded area depicts a 95 % confidence interval about the measured value. Samples at which the confidence area does not include 0 correspond to a rejection of the hypothesis at the uncorrected .05 criterion. For all tests, pale marks at y = 0 indicate samples at which the hypothesis is rejected at the uncorrected .05 criterion, and dark marks indicate samples at which the hypothesis is rejected after FDR correction. All tests were done with N = 13 subjects
Mentions: This experiment used EEG data from 13 subjects performing a target detection task in a rapid serial visual presentation paradigm. After artifact rejection, an average of 48.5 (range 25–54) target trials and 426.1 (range 240–496) non-target trials were maintained. In the terms used in Experiments 1 and 2, data imbalance as the proportion of target trials was 1/9.8 on average (range 1/8.8–1/10.6). The three statistical tests (unbalanced paired permutation, conventional paired-samples permutation, paired-samples T test) were applied to the GFP at each time point from 1000 ms before stimulus onset to 2000 ms after onset. At the sampling rate of 256 Hz, this results in 768 time points. Results of the four statistical tests are shown in Fig. 4. All three tests rejected their respective hypotheses (q < .05, FDR corrected) from 296 to 1051 ms post-stimulus onset. However, consistent with the simulation results from Experiment 1, the conventional statistical tests rejected their respective hypotheses prior to stimulus onset 239 and 235 times out of 256 total pre-stimulus samples for paired-samples T test and paired-samples permutation test, respectively.Fig. 4

Bottom Line: Global field power is a valuable summary of multi-channel electroencephalography data.The results show that the proposed test finds global field power differences in the classical P3 range; the other tests find differences in that range but also at other times including at times before stimulus onset.These results are interpreted as showing that the proposed test is valid and sensitive to real within-subject differences in global field power in multi-subject unbalanced data.

View Article: PubMed Central - PubMed

Affiliation: U.S. Army Research Laboratory, Aberdeen Proving Ground, MD, USA. benjamin.t.files.civ@mail.mil.

ABSTRACT
Global field power is a valuable summary of multi-channel electroencephalography data. However, global field power is biased by the noise typical of electroencephalography experiments, so comparisons of global field power on data with unequal noise are invalid. Here, we demonstrate the relationship between the number of trials that contribute to a global field power measure and the expected value of that global field power measure. We also introduce a statistical testing procedure that can be used for multi-subject, repeated-measures (also called within-subjects) comparisons of global field power when the number of trials per condition is unequal across conditions. Simulations demonstrate the effect of unequal trial numbers on global field power comparisons and show the validity of the proposed test in contrast to conventional approaches. Finally, the proposed test and two alternative tests are applied to data collected in a rapid serial visual presentation target detection experiment. The results show that the proposed test finds global field power differences in the classical P3 range; the other tests find differences in that range but also at other times including at times before stimulus onset. These results are interpreted as showing that the proposed test is valid and sensitive to real within-subject differences in global field power in multi-subject unbalanced data.

No MeSH data available.


Related in: MedlinePlus