Here you can see the ‘live results’ of the trial as they come in, including the number of participants in each group, and how this is broken down by age and sex (as the participants are allocated randomly to the different groups this should be more or less equal, but confirming this is important). You can also see a summary of the error rates between the groups, however we are keeping the identities of the study group concealed until we have exceeded the required sample size (why? To avoid people who have not yet done the study getting the impression that one type of music or another is going to improve or detract from their performance!). After we have recruited sufficient participants we will reveal the identity of the groups, as well as run the full analysis as stated in the Statistical Plan to determine whether or not our hypothesis was correct.
To receive the final results and conclusions once they are in please sign up to Demos.science at the end of this exercise.
Table 1: Summary of Demographic characteristics
|Parameters||Grouped by group|
|number of participants*||19||27||20|
|gender, number (%)||Male||7 (36.8%)||5 (18.5%)||5 (25.0%)|
|Female||10 (52.6%)||18 (66.7%)||14 (70.0%)|
|Other/not provided||2 (10.5%)||4 (14.8%)||1 (5.0%)|
* 4 participants who failed to tap the screen more than 5 times in a row were excluded from the analysis, as they likely misread the instructions or had network issues. This affected 4 participants in Group X, 0 participants in Group Y and 0 participants in Group Z.
The graph below shows the average number of error rates in each group (for now you’ll have to guess which is which!). It’s critical to remember however that this average, or any differences in averages between groups might not be at all precise, or indicate that there is a real difference between the groups. This will depend to a great extent on the number of participants in the group – if it is small, then any differences we’re seeing could be entirely due to chance.
Fig 1: Distribution of error rates across three groups.
The brackets at the top of each bar are there to give us an indication of how precise the average is. If we have a small number of participants (and if their performance on the test is wildly different) then this confidence interval will be very wide, suggesting that the average so far is not at all precise. On the graph this will show up as having wide 'confidence intervals', and they are likely to be overlapping when comparing them across the groups. As more participants take part the confidence intervals are likely to narrow down. Eventually, we might find that the confidence intervals no longer overlap, and this would suggest that the differences between the groups are genuine, and not just due to chance (i.e. 'statistically significant' differences).
The notes above the graph summarize all this by indicating whether or not there is a statistically significant difference between the groups or whether any apparent difference is likely to be due to chance (not significant).
In the remaining two graphs we look at the differences in SART error rates between participants of different age groups and sex, regardless of the group they were allocated to. In the final analysis we will carry out further ‘subgroup analyses’ to see if there were differences in the impact of music on concentration between these subgroups.
Fig 2: Distribution of error rates across genders
Fig 3: Distribution of error rates across age groups