Descriptive statics
The results related to mean, standard deviation, skewness and stretching of five scales HR, RR, LF, HF, LFtoHF-TASK are shown in Table 1. As can be seen, the values of skewness and elongation of HR, skewness in RR and skewness and elongation in LFtoHF scale are less than (16), therefore, citing Tabachinek and Fidel (1996), it can be concluded that the distribution of scores of the three scales is relatively normal. and they have no problem in terms of curvature and tension.
Table 1
The results of descriptive statistics for five scales HR, RR, LF, HF, LftoHF
Scales | Mean | SD | Skeweness | Kurtosis |
HR | 81.46 | 9.32 | -0.19 | -0.75 |
RR | 754.90 | 108.04 | -1.22 | 7.15 |
LF | 1592.78 | 3017.16 | 4.05 | 17.24 |
HF | 3154.59 | 5651.46 | 2.68 | 6.63 |
LFTOHF | 0.79 | 0.56 | 1.49 | 1.88 |
The results of the t test are shown in Table 2. Independent groups t-test was used to examine the difference between two groups of high anger and low anger in each scale of HR, RR, LF, HF, LFtoHF.
Table 2
The results of t test (t) of independent groups
خطای استاندارد برآورد تفاوت | تفاوت میانگین ها | سطح معنی داری | t | df | مقیاس ها |
17/2 | 57/5 | 012/0 | 56/2 | 74 | HR |
83/24 | 40/74- | 004/0 | 99/2- | 74 | RR |
45/725 | 35/984- | 17/0 | 35/1- | 74 | LF |
14/1357 | 19/1935- | 15/0 | 42/1- | 74 | HF |
13/0 | 17/0 | 19/0 | 29/1 | 74 | LFtoHF |
As can be seen, there is a significant difference between the scores of the two groups in the HR and RR scales. However, no significant difference was observed in the other three scales. The resulting t-statistic about the HR scale was statistically significant (p < 0.05 and t = (76) 2.56). Therefore, there is a significant difference between the two groups of high anger and low anger in this scale. The resulting t-statistic about the RR scale was statistically significant (p < 0.05 and t = -2.99 (76)). Therefore, there is a significant difference between the two groups of high anger and low anger in this scale.
d) Examining the diagnostic accuracy of the scales
ROC analysis and sensitivity analysis were used to investigate the three scales HR, RR, LF, HF, and LFtoHF. The results of ROC analysis are presented in Table 3. As it is clear in the table, the RR scale has the highest value of cleanness index ( ) compared to other scales (= 0.71) and after that the HR scale has the highest value of cleanness index (= 0.69). But the RR scale has a significantly better performance than the other scale, namely HR (Z = 0.01 and Z = 0.06). This neat index shows that the RR scale is able to correctly classify people into two groups of people with high aggression and low aggression in 71% of cases. Although the HR scale did not obtain a higher purity index (0.69), it has a better performance than other scales (Z = 0.06 and 0.01). The rest of the scales i.e. HF, LF, LFtoHF do not have good performance and clinical value.
Table 3
The results of ROC analysis for two groups of people with high and low aggression
Scales | AZ | SE | CI |
RR | 0.71 | 0.06 | 0.60–0.81 |
HR | 0.69 | 0.06 | 0.57–0.79 |
HF | 0.59 | 0.07 | 0.47–0.70 |
LF | 0.54 | 0.06 | 0.42–0.66 |
LFtoHF | 0.55 | 0.07 | 0.43–0.66 |
As can be seen in Table 3, the level under the curve of the RR scale is 71%, as mentioned above. The only scale that can identify the emotion of anger in people with a probability of 71% and with a confidence interval (0.81 − 0.60) (P = 0.001). In the following chart, this clean index is shown in more detail and in comparison with other scales.
For this graph, we chose the categorized variable of anger and all the heart rate variables including HR-TASK, RR-TASK, LF-TASK, HF-TASK, LFtoHF-TASK and rock analysis showed that RR-TASK The rest of the variables are more significant.Therefore, we check the RR scale separately in the chart below, and its values are also reported.
As seen in Fig. 2, the area under the ROC curve for the RR scale is 71% (P = 0.001). This means that the heart rate RR scale can detect anger in highly aggressive individuals with a probability of 71%. The sensitivity and specificity are also explained in the following tables.
In order to determine the optimal cut score for five scales RR, HR, HF, LF, LFtoHF, sensitivity analysis was used. In sensitivity analysis, several characteristics are considered, which include sensitivity, specificity, positive likelihood ratio and negative likelihood ratio. Of course, sometimes, instead of positive probability ratio and negative probability ratio, two indicators of false positive rate and false negative rate are considered. A suitable cut-off score is a test score that can optimize the value of the mentioned 4 characteristics. Among the 4 mentioned characteristics, sensitivity is very important. The optimal cut-off score is a score that, while keeping the sensitivity and specificity values at their maximum value, keeps the negative probability ratio at its minimum value. Therefore, in a wide range of cut-off points (examination scores), only one score represents the desired cut-off score that can maintain the mentioned state. It is also important to mention this point that regardless of the cut-off score provided, this cut-off score is effective when the scale has a good purity index .
Table 4
Area under the ROC curve for the RR scale
Area under the ROC curve (AUC) | SE | CI | Z | sig |
0.71 | 0.06 | 0.60–0.81 | 3.20 | 0.001 |
جدول 5: شاخص Youden برای مقیاس RR
Youden index | Cut point | Sensivity score | Characteristic score |
0.45 | 88 | 66/690< | 0.45 |
As stated before, in the current research, only RR has a favorable and relatively good cleanliness index. Therefore, according to the total of two ROC and sensitivity analyses, it can be concluded that if a person scores higher than 690.66 in the RR scale, there is a 71% probability that he has anger and aggressive behaviors. These results are favorable compared to previous similar studies and this shows that the use of a heart rate sensor can be a suitable and reliable tool to detect and identify the emotion of anger.
In order to determine the optimal cutoff, we are faced with an index called the Youden index. Based on this index, our optimal and desirable cut-off point is the one with the highest sum of sensitivity and specificity. The accuracy in the above results shows that the number 690/66 with sensitivity equal to 88.00% can be considered as the best cut point.