Step 4a: Processing Practice VHT Data
Practice phase data are processed for each individual to classify the main VHT responses as hits and false alarms.
i) Locating necessary information:
Open a practice edata file (data file produced by the E-Prime software) and locate the column named ‘RespTime’ in it.
ii) Responses in RespTime columns except RespTime 1 are false alarms:
‘RespTime’ column contains reaction time (i.e. how long it took for a participant to respond to a voice; see Table 3) in milliseconds to each sound file. ‘NULL’ is a miss trial meaning that there was a voice but participants missed it (i.e. failed to respond).
RespTime1 column contains reaction time data which are used to define the criteria for hits and false alarms in the main VHT
During the practice, participants usually respond once to each voice stimulus and this produces a single “RespTime” column (e.g. RespTime 1 in Table 3) but for some participants there might be more than one column as Table 4 shows. Responses in these additional columns (e.g. RespTime2 in Table 4) are regarded as false alarms and are excluded from practice data (i.e. are not processed further).
Reaction time values in RespTime columns other than RespTime1 are false alarms
They are false alarms because participants are instructed to respond once to each voice stimulus, so multiple responses mean that participants responded not because they heard a voice but because they thought that there was a voice. As Table 4 demonstrates, the participant took 251 ms to respond to the last voice and it was likely to be a hit, but then s(he) gave another response after 2518 ms. This second reaction time clearly does not correspond to his/her general reaction time pattern and there was no immediate voice stimulus that preceded it so it can only be described as a false alarm.
iii) Identifying false alarms in RespTime1 column:
Above threshold sound intensities are used for more than half of the sound files in the practice VHT, so it is expected that majority of the responses would represent participants’ true auditory reaction time. But it is possible that the immediate response to a sound file actually is a false alarm. One way of identifying this is to look for response times that are either too short or too long (i.e. outliers) and thus lie outside the overall pattern of reaction time distribution. It is expected that the remaining reaction time values (after excluding outliers) would sufficiently be accurate approximation to actual reaction time. An outlier-labelling rule (Hoaglin & Lglewicz, 1987) is used to identify and exclude the outliers. The use of this rule is demonstrated using the data presented in Table 3.
Following are the steps of detecting outliers:
First, the lower quartile (Q1 or first quartile, i.e. the median of the lower half of the data set) and the upper quartile (Q3 or third quartile, i.e. the median of the upper half of the data set) are determined.
In the example,
Q1 = 405
Q3 = 824
Second, the difference between the upper quartile and the lower quartile is determined.
Q3 – Q1 = 419
Third, Q3 – Q1 value is multiplied by a factor called g’. The value of g is 2.2 (Hoaglin & Lglewicz, 1987).
g' = 419 x 2.2 = 921.8
Fourth, the value of g' is subtracted from Q1 and the same value is added to Q3 to obtain lower and upper limits respectively. In the example, the lower limit is -516.8 and the upper limit is 1745.8, so an outlier in the practice VHT would be any reaction time which falls outside the range of -516.8 to 1745.8 (i.e. lower than -516.8ms or higher than 1745.8ms). In our example, all the values are retained for the next step (to determine the mean and SD of RespTime1 values) as they are within this range.
iv) Determining the mean and SD of RespTime1 values:
Mean and standard deviation of the practice response time are determined after excluding the outliers (if there are any). In the example,
Mean response time = 597.73
SD = 274.47
v) Defining false alarms for the main VHT:
Mean and standard deviation values as determined in the previous step are used to identify false alarms in the main VHT, which are defined as any reaction time below or above 2SD from the mean as obtained in the practice VHT. Thus, in our example, reaction time values in the main VHT (Table 5; see the section on how to process the main VHT data) shorter than 49ms (597.73-2*274.47) or longer than 1147ms (597.73+2*274.47) are labelled as false alarms.
Note. Given that participants are instructed to respond to voice stimuli as quickly as possible, it is believed that all their hits should remain within two standard deviations above and below the mean response time, as in a normal distribution 95% of the values lie within two standard deviations of the mean. But the standard deviation criterion is not fixed—one can leniently define a hit as the reaction time that lies within three standard deviations from the mean whereas it can stringently be defined as any response time within one standard deviation from the mean.
Hoaglin, D. C., & Iglewicz, B. (1987). Fine-tuning some resistant rules for outlier labeling. Journal of the American Statistical Association, 82(400), 1147-1149.