Data Reduction

Question 1

You mentioned excluding TCD data affected by signal aliasing or loss. What criteria did you use to identify and exclude such data points?

Answer 1

We visually inspected the TCD waveforms to identify any abnormalities or artifacts (i.e., drop-offs due to sudden head movements or values that were more/less than a 60% increase/decrease from baseline). Outliers and cut-offs were evaluated on a case-by-case bases due to variable baseline MCAv and variations in levels of sensitivity to exercise.

Question 2

Why did you choose to examine peak systolic MCAv as a proxy for exercise-mediated changes in CBF? Are there any limitations to using this measure?

Answer 2

• Peak systolic MCAv is a commonly used measure in TCD studies to assess changes in cerebral blood flow velocity. It is considered a reliable indicator of changes in CBF, as it reflects the maximum velocity of blood flow during the systolic phase of the cardiac cycle. However, there are some limitations to using peak systolic MCAv as a proxy for exercise-mediated changes in CBF:
• a) MCAv is a measure of blood flow velocity, not actual blood flow. Changes in vessel diameter can influence MCAv without necessarily reflecting changes in CBF.
• b) TCD measures blood flow velocity in the middle cerebral artery, which supplies a specific region of the brain. It may not accurately represent global changes in CBF.
  c) Factors such as individual differences in vessel anatomy, probe placement, and the angle of insonation can affect the accuracy of MCAv measurements.

Question 3

For the oculomotor task, you used a dual-pass Butterworth filter with a low-pass cut-off frequency of 15 Hz. How did you determine this cut-off frequency, and what are the implications of using this specific filter?

Answer 3

• A low-pass cut-off frequency of 15 Hz would remove high-frequency noise above 15 Hz, which may include artifacts related to muscle activity or equipment noise.
• The implications of using this specific filter are that it may remove some high-frequency components of the eye movement signal, potentially affecting the accuracy of saccade detection and measurement.
  However, if the relevant saccade information is contained within the frequency range below 15 Hz, this filter should preserve the important features of the signal for analysis.

Question 4

You used split-plot ANOVAs for analyzing MCAv, HR, and oculomotor dependent variables. Can you explain the rationale behind choosing this statistical approach?

Answer 4

• Split-plot ANOVAs, also known as mixed-design ANOVAs, are used when there are both between-subjects and within-subjects factors in the study design.
• In this case, the between-subjects factor was the group (SRC vs. HC), and the within-subjects factors were time (pre- vs. postexercise) and task (prosaccade vs. antisaccade for the oculomotor variables).
• The split-plot ANOVA allows for the examination of main effects and interactions between these factors, making it an appropriate choice for analyzing the MCAv, HR, and oculomotor data in this study.
  This approach takes into account the repeated measures nature of the data and allows for the comparison of changes over time between the two groups.

Question 5

How did you handle violations of sphericity in your data, and what are the implications of using the Huynh-Feldt correction?

Answer 5

• Sphericity is an assumption of repeated measures ANOVA that requires equal variances of the differences between all pairs of the within-subjects conditions.
• When this assumption is violated, the F-statistic can be positively biased, increasing the risk of Type I errors.
• To handle violations of sphericity, we used the Huynh-Feldt correction, which adjusts the degrees of freedom of the F-test to account for the degree of sphericity violation.
• The Huynh-Feldt correction is more conservative than the Greenhouse-Geisser correction and is recommended when the epsilon value (a measure of sphericity) is greater than 0.75.
  By using this correction, we have taken steps to control for the potential bias introduced by sphericity violations and maintain the validity of their statistical inferences.

Question 6

You used planned comparison paired-sample t-tests for SCAT-5 symptom frequency and severity. Why did you choose this approach instead of including these variables in the split-plot ANOVA?

Answer 6

• We chose to use planned comparison paired-sample t-tests for the SCAT-5 symptom frequency and severity data because these variables were only assessed in the SRC group and at different time points than the other dependent variables.
• Specifically, SCAT-5 data were collected at pre-BCBT (Visit 1), postexercise oculomotor assessment (end of Visit 2), and 24-hour follow-up, whereas the other variables were measured pre- and postexercise within Visit 2.
• As a result, including the SCAT-5 data in the split-plot ANOVA would have been inappropriate due to the different time points and the absence of a between-subjects factor (since the HC group did not complete the SCAT-5).
  Using planned comparison paired-sample t-tests allowed us to make specific comparisons of interest (e.g., pre-BCBT vs. postexercise, pre-BCBT vs. 24-hour follow-up) within the SRC group, without the need to include a between-subjects factor or align the time points with the other dependent variables.

Question 7

Justify your choice of recruiting 16 participants per group.

Answer 7

• The decision to recruit 16 participants per group was based on an a priori power analysis conducted using G*Power.
• G*Power analysis (“means: differences between two independent means.” (power =0.90, alpha = 0.05) from an effect size (Cohen’s d = 1.22) derived from Ayala and Heath’s (2020) comparison of antisaccade reaction times between persons in acute SRC recovery and age- and sex-matched healthy controls. The analysis yielded a sample size of 16 participants per group.
• An a priori power analysis is a statistical method used to determine the sample size required to detect an effect of a specified size with a desired level of statistical power and significance before conducting a study.
  This approach helps to ensure that the study is adequately powered to detect meaningful differences between groups, reducing the risk of Type II errors (false negatives).
  This approach helps to ensure that the study is adequately powered to detect meaningful differences between groups, reducing the risk of Type II errors (false negatives).

Question 8

In this study, the a priori power analysis was performed with the following parameters:

Answer 8

• The goal was to detect differences between two independent means, specifically between the SRC and HC groups.
• The desired statistical power was set at 0.90, which means there is a 90% probability of detecting a true effect if one exists. This high level of power was chosen to minimize the risk of Type II errors.
• The significance level (alpha) was set at 0.05, which is a commonly used threshold for determining statistical significance.
• The effect size (Cohen’s d) was set at 1.22, based on the findings of a previous study by Ayala and Heath (2020).
• Cohen’s d is a standardized measure of the difference between two means, with values of 0.2, 0.5, and 0.8 generally considered small, medium, and large effects, respectively.
• The effect size of 1.22 obtained from Ayala and Heath (2020) suggests a large difference between the groups.
• Based on these parameters, the a priori power analysis indicated that a sample size of 16 participants per group would be sufficient to detect the expected effect size with the desired level of statistical power and significance.
• This sample size determination helps to optimize the study design by balancing the need for sufficient statistical power with the practical constraints of participant recruitment and study resources.

Question 9

Which assumption tests did you consider for each analysis:

Answer 9

• 1. Split-plot ANOVAs (MCAv, HR, and oculomotor data):
• Independence of observations: Ensure that the observations within each group are independent of each other.
• Normality: Check if the dependent variables are normally distributed within each group using tests like Shapiro-Wilk or Kolmogorov-Smirnov, or by examining Q-Q plots and histograms.
• Homogeneity of variances: Assess whether the variances of the dependent variables are equal across groups using Levene’s test.
• Sphericity: For within-subjects factors with more than two levels, check if the variances of the differences between all pairs of levels are equal using Mauchly’s test of sphericity. If violated, apply the Huynh-Feldt correction to the degrees of freedom.
• 2. Paired-sample t-tests (SCAT-5 symptom frequency and severity):
• Normality: Verify if the differences between the paired observations are normally distributed using tests like Shapiro-Wilk or Kolmogorov-Smirnov, or by examining Q-Q plots and histograms.
• Independence of observations: Ensure that the pairs of observations are independent of each other.
• 3. Pearson’s r (MCAv, HR, and antisaccade RTs):
• Linearity: Check if there is a linear relationship between the variables using scatterplots.
• Normality: Assess if the variables are normally distributed using tests like Shapiro-Wilk or Kolmogorov-Smirnov, or by examining Q-Q plots and histograms.
• Homoscedasticity: Verify if the variability in one variable is similar across all values of the other variable using scatterplots.
• Independence of observations: Ensure that the observations are independent of each other.
• If any assumptions are violated, consider applying appropriate transformations to the data or using alternative non-parametric tests. For example, if normality is violated, you may use the Wilcoxon signed-rank test instead of a paired-sample t-test or Spearman’s rank correlation instead of Pearson’s r.
• When reporting the results, mention that you checked the relevant assumptions for each test and whether any violations were observed. If violations were found, describe how you addressed them (e.g., applying transformations or using alternative tests).