![An ANCOVA was performed on the preFinal percentage and postFinal percentage using a fixed
factor of Absences and a covariate of missing assignments. The results showed the covariate,
missing assignments was significantly related to absences F(1, 23) = 17.07, p < .05, r = .43.
Also it must be noted that there was not a significant effect on of absences on postFinal
percentage after controlling the effect of missing assignments F(4, 23) = 2.85, p < .05, partial η2
= .33 (see Tables 2 through 4).
Table 2
a
Levene's Test of Equality of Error Variances
Dependent Variable:Post-Final Percentage
F df1 df2 Sig.
1.504 4 24 .233
Tests the null hypothesis that the error variance
of the dependent variable is equal across groups.
a. Design: Intercept + missing + absences
Table 3
Tests of Between-Subjects Effects
Dependent Variable:Post-Final Percentage
Source Type III Sum of
Squares df Mean Square F Sig.
a
Corrected Model 3346.783 5 669.357 13.211 .000
Intercept 21510.189 1 21510.189 424.542 .000
missing 864.841 1 864.841 17.069 .000
absences 577.985 4 144.496 2.852 .047
Error 1165.335 23 50.667
Total 206573.790 29
Corrected Total 4512.118 28
a. R Squared = .742 (Adjusted R Squared = .686)
Table 4
Parameter Estimates
Dependent Variable:Post-Final Percentage
Parameter 95% Confidence Interval
B Std. Error t Sig. Lower Bound Upper Bound
Intercept 84.068 11.078 7.589 .000 61.152 106.984
missing -.596 .144 -4.131 .000 -.895 -.298
[absences=0] 15.984 9.770 1.636 .115 -4.225 36.194
[absences=1] 5.915 9.407 .629 .536 -13.545 25.375
[absences=2] 8.012 10.273 .780 .443 -13.239 29.263](https://image.slidesharecdn.com/gradestatistics-1284639608-phpapp02/85/Grade-Statistics-4-320.jpg)
![[absences=3] 10.046 9.024 1.113 .277 -8.620 28.713
a
[absences=5] 0 . . . . .
a. This parameter is set to zero because it is redundant.
The results further showed that there was not a significant effect on of absences on preFinal
percentage after controlling the effect of missing assignments F(4, 23) = 2.90, p < .05, partial η2
= .34 (see Tables 5 through 7).
Table 5
a
Levene's Test of Equality of Error Variances
Dependent Variable:Pre-Final Percentage
F df1 df2 Sig.
.832 4 24 .518
Tests the null hypothesis that the error variance
of the dependent variable is equal across groups.
a. Design: Intercept + missing + absences
Table 6
Tests of Between-Subjects Effects
Dependent Variable:Pre-Final Percentage
Source Type III Sum of
Squares df Mean Square F Sig.
a
Corrected Model 4417.947 5 883.589 15.376 .000
Intercept 23406.178 1 23406.178 407.303 .000
missing 1235.160 1 1235.160 21.494 .000
absences 666.504 4 166.626 2.900 .044
Error 1321.723 23 57.466
Total 217754.020 29
Corrected Total 5739.670 28
a. R Squared = .770 (Adjusted R Squared = .720)
Table 7
Parameter Estimates
Dependent Variable:Pre-Final Percentage
Parameter 95% Confidence Interval
B Std. Error t Sig. Lower Bound Upper Bound
Intercept 87.709 11.797 7.435 .000 63.304 112.114
missing -.713 .154 -4.636 .000 -1.031 -.395
[absences=0] 17.031 10.404 1.637 .115 -4.492 38.555](https://image.slidesharecdn.com/gradestatistics-1284639608-phpapp02/85/Grade-Statistics-5-320.jpg)
![[absences=1] 6.668 10.018 .666 .512 -14.057 27.392
[absences=2] 7.402 10.940 .677 .505 -15.230 30.034
[absences=3] 10.512 9.610 1.094 .285 -9.368 30.391
a
[absences=5] 0 . . . . .
a. This parameter is set to zero because it is redundant.
Lastly a correlation was performed on all variables (see Table 8) in order to find relationships
between the variables. The variable gender showed no significance to our test and will be
excluded from this report; however, all other variables showed some level of significant
correlation with other variables at the p < .01 level and the findings will be reported here.
Absences showed a significant correlation to missing assignments r = .59, preFinal percentage r
= -.67, and postFinal percentage r = -.65, (all ps < .01).
Missing assignments showed a significant correlation to preFinal r = -.81 and postFinal r = -.78,
(all ps < .01).
The preFinal percentage showed a significant correlation to postFinal r = .97, p < .01, and the
final percentage only showed a significant correlation with postFinal r = .50, p < .01.](https://image.slidesharecdn.com/gradestatistics-1284639608-phpapp02/85/Grade-Statistics-6-320.jpg)
Grade Statistics
- 1. Grade statistics from two random classes Introduction The classes used in this study (Class A and Class B) were held at the same campus in Wichita, KS during June through September 2010 by the same instructor. These two classes were held on a Thursday and a Friday night with Class A being held on Thursday and Class B being held on Friday. Class A completed with 13 students and Class B completed with 16 students. The following analysis was performed by extracting statistics based upon the following variables. Class = a numeric value assigned to each class starting with A1 being 1 through A6 being 11 thus for Thursday and Friday evening classes the data would be 8 and 10 respectively Absences = a numeric value for the number of classes a given student missed during the quarter Missing = the missing assignment percentage for a given student. For Class A there were 34 assignments and for Class B there were 25. Gender = a numeric value representing the gender of the student 0 = Male and 1 = Female preFinal = the percentage the student had going into the final exam Final = the percentage earned on the final exam postFinal = the final percentage earned in the class Background The most interesting thing about these two groups of students is that the one group was overprotected through most classes leading up to the class in question and their general attitudes during this class reflected their attitudes prior to this specific class, while the other group was a group of first term students. These first term students were told up front what was expected of them and little to no tolerance was given for late work submission (this rule was also applied to the group that had been overprotected prior to this class). The first term group in turn appeared to be in attendance more frequently, appeared to turn in work with more regularity, and appeared to be less needy than the more experienced students. There were three drops from the new students, but the absentee rate (absences versus 11 weeks * number of students) for the students that completed the quarter was 8% (14/176) compared to the second group’s 11% (16/143) rate. However, it should be noted that the second group was hurt by the one failure this quarter, who alone accounted for five of the 16 absences, as this student regularly missed two classes and was then present for one class. Removing these five absences from their rate takes the group absentee rate down to 8% (11/143) as well. If you add in the dropped student absentee rate to the new students total you get an additional 11 absentees out of 26 classes or 25/212 which is 12%. It should be noted that, of these three drops two missed the first night of class without notifying the school, but were able to catch up; and all three had good grades going into their string of absences that ultimately led to their drop from school. Testing and Analysis After the data was entered tests were conducted on the variables pre/postFinal and final in an attempt to analyze if the data was normally distributed.
- 2. Figure 1 Figure 2 Figure 3 What was noted from this (see figures 1-3), was that the data appeared slightly skewed, but otherwise fairly normal. Table 1 Statistics Pre-Final Final Post-Final Percentage Percentage Percentage N Valid 29 29 29 Missing 0 0 0 Mean 85.5034 75.3103 83.4724 Std. Error of Mean 2.65868 2.92480 2.35729 Median 83.7000 76.0000 82.2000 a Mode 73.10 92.00 85.90 Std. Deviation 14.31741 15.75052 12.69437
- 3. Variance 204.988 248.079 161.147 Skewness -.413 -.102 -.307 Std. Error of Skewness .434 .434 .434 Kurtosis .552 -1.029 .686 Std. Error of Kurtosis .845 .845 .845 Percentiles 25 76.3500 62.5000 75.2000 50 83.7000 76.0000 82.2000 75 95.9000 92.0000 93.1000 a. Multiple modes exist. The smallest value is shown Further analysis (see Table 1) showed that all variables were within tolerance and that the only variable showing any signs of abnormality regarded the kurtosis of the final percentage - 1.029/.845 = -1.22, but it was still within the tolerance of 1.96. So it was concluded that all data was normally distributed. The mean pre-final percentage grade did appear to decrease with the number of absences (see Figure 4). Additionally the mean pre-final percentage appeared to decrease with the number of missing assignments (see Figure 5). Figure 4 Figure 5
- 4. An ANCOVA was performed on the preFinal percentage and postFinal percentage using a fixed factor of Absences and a covariate of missing assignments. The results showed the covariate, missing assignments was significantly related to absences F(1, 23) = 17.07, p < .05, r = .43. Also it must be noted that there was not a significant effect on of absences on postFinal percentage after controlling the effect of missing assignments F(4, 23) = 2.85, p < .05, partial η2 = .33 (see Tables 2 through 4). Table 2 a Levene's Test of Equality of Error Variances Dependent Variable:Post-Final Percentage F df1 df2 Sig. 1.504 4 24 .233 Tests the null hypothesis that the error variance of the dependent variable is equal across groups. a. Design: Intercept + missing + absences Table 3 Tests of Between-Subjects Effects Dependent Variable:Post-Final Percentage Source Type III Sum of Squares df Mean Square F Sig. a Corrected Model 3346.783 5 669.357 13.211 .000 Intercept 21510.189 1 21510.189 424.542 .000 missing 864.841 1 864.841 17.069 .000 absences 577.985 4 144.496 2.852 .047 Error 1165.335 23 50.667 Total 206573.790 29 Corrected Total 4512.118 28 a. R Squared = .742 (Adjusted R Squared = .686) Table 4 Parameter Estimates Dependent Variable:Post-Final Percentage Parameter 95% Confidence Interval B Std. Error t Sig. Lower Bound Upper Bound Intercept 84.068 11.078 7.589 .000 61.152 106.984 missing -.596 .144 -4.131 .000 -.895 -.298 [absences=0] 15.984 9.770 1.636 .115 -4.225 36.194 [absences=1] 5.915 9.407 .629 .536 -13.545 25.375 [absences=2] 8.012 10.273 .780 .443 -13.239 29.263
- 5. [absences=3] 10.046 9.024 1.113 .277 -8.620 28.713 a [absences=5] 0 . . . . . a. This parameter is set to zero because it is redundant. The results further showed that there was not a significant effect on of absences on preFinal percentage after controlling the effect of missing assignments F(4, 23) = 2.90, p < .05, partial η2 = .34 (see Tables 5 through 7). Table 5 a Levene's Test of Equality of Error Variances Dependent Variable:Pre-Final Percentage F df1 df2 Sig. .832 4 24 .518 Tests the null hypothesis that the error variance of the dependent variable is equal across groups. a. Design: Intercept + missing + absences Table 6 Tests of Between-Subjects Effects Dependent Variable:Pre-Final Percentage Source Type III Sum of Squares df Mean Square F Sig. a Corrected Model 4417.947 5 883.589 15.376 .000 Intercept 23406.178 1 23406.178 407.303 .000 missing 1235.160 1 1235.160 21.494 .000 absences 666.504 4 166.626 2.900 .044 Error 1321.723 23 57.466 Total 217754.020 29 Corrected Total 5739.670 28 a. R Squared = .770 (Adjusted R Squared = .720) Table 7 Parameter Estimates Dependent Variable:Pre-Final Percentage Parameter 95% Confidence Interval B Std. Error t Sig. Lower Bound Upper Bound Intercept 87.709 11.797 7.435 .000 63.304 112.114 missing -.713 .154 -4.636 .000 -1.031 -.395 [absences=0] 17.031 10.404 1.637 .115 -4.492 38.555
- 6. [absences=1] 6.668 10.018 .666 .512 -14.057 27.392 [absences=2] 7.402 10.940 .677 .505 -15.230 30.034 [absences=3] 10.512 9.610 1.094 .285 -9.368 30.391 a [absences=5] 0 . . . . . a. This parameter is set to zero because it is redundant. Lastly a correlation was performed on all variables (see Table 8) in order to find relationships between the variables. The variable gender showed no significance to our test and will be excluded from this report; however, all other variables showed some level of significant correlation with other variables at the p < .01 level and the findings will be reported here. Absences showed a significant correlation to missing assignments r = .59, preFinal percentage r = -.67, and postFinal percentage r = -.65, (all ps < .01). Missing assignments showed a significant correlation to preFinal r = -.81 and postFinal r = -.78, (all ps < .01). The preFinal percentage showed a significant correlation to postFinal r = .97, p < .01, and the final percentage only showed a significant correlation with postFinal r = .50, p < .01.
- 7. Table 8 Correlations Missing Assignment Pre-Final Final Post-Final Absences Percentage Gender Percentage Percentage Percentage ** ** ** Absences Pearson Correlation 1 .593 .061 -.670 -.190 -.652 Sig. (2-tailed) .001 .755 .000 .323 .000 Sum of Squares and Cross- 44.966 280.455 .828 -340.203 -106.310 -293.672 products Covariance 1.606 10.016 .030 -12.150 -3.797 -10.488 N 29 29 29 29 29 29 ** ** ** Missing Assignment Pearson Correlation .593 1 .081 -.808 -.216 -.783 Percentage Sig. (2-tailed) .001 .677 .000 .260 .000 Sum of Squares and Cross- 280.455 4977.432 11.576 -4321.174 -1272.103 -3712.344 products Covariance 10.016 177.765 .413 -154.328 -45.432 -132.584 N 29 29 29 29 29 29 Gender Pearson Correlation .061 .081 1 -.266 .044 -.229 Sig. (2-tailed) .755 .677 .164 .821 .233 Sum of Squares and Cross- .828 11.576 4.138 -40.917 7.448 -31.262 products Covariance .030 .413 .148 -1.461 .266 -1.117 N 29 29 29 29 29 29 ** ** ** Pre-Final Percentage Pearson Correlation -.670 -.808 -.266 1 .275 .971 Sig. (2-tailed) .000 .000 .164 .150 .000 Sum of Squares and Cross- -340.203 -4321.174 -40.917 5739.670 1733.469 4941.903 products
- 8. Covariance -12.150 -154.328 -1.461 204.988 61.910 176.497 N 29 29 29 29 29 29 ** Final Percentage Pearson Correlation -.190 -.216 .044 .275 1 .496 Sig. (2-tailed) .323 .260 .821 .150 .006 Sum of Squares and Cross- -106.310 -1272.103 7.448 1733.469 6946.207 2777.448 products Covariance -3.797 -45.432 .266 61.910 248.079 99.195 N 29 29 29 29 29 29 ** ** ** ** Post-Final Percentage Pearson Correlation -.652 -.783 -.229 .971 .496 1 Sig. (2-tailed) .000 .000 .233 .000 .006 Sum of Squares and Cross- -293.672 -3712.344 -31.262 4941.903 2777.448 4512.118 products Covariance -10.488 -132.584 -1.117 176.497 99.195 161.147 N 29 29 29 29 29 29 **. Correlation is significant at the 0.01 level (2-tailed).
- 9. Conclusions What seems clear from this analysis is that although there is a correlation between absenteeism and final grades, the overriding factor appears to be missing assignments. While absenteeism played a role in the participant’s grade, it would seem that absenteeism correlated more directly with drop rates, and missing assignments had a more significant effect on both preFinal and postFinal scores than did attendance alone. Also, while the final exam did have a significant correlation to the participant’s postFinal score, as would be expected; the preFinal percentage did not play a significant role in the final exam score. This appears to indicate that although a student misses class on a regular basis, they can still score well on the final exam, as would be expected if they study. However, if they continually fail to attend class and fail to turn in assignments, this has little impact on the final exam score, but it does negatively impact preFinal scores which cannot be overcome by a high final exam score. It appears that attendance has a limited impact on students’ scores, but the overpowering negative that affects student success is failure to turn in assignments. This in turn places them at a severe disadvantage prior to the final. The question is, how much is a faculty member able to influence attendance? While there is some impact due to likeability, style, etc., that we have not addressed with this study, there is nothing that a faculty member can do to force attendance; and what this study clearly shows is that even if a student attends, failure to submit work will still doom them to failure.