I was just wondering if I was interpreting my findings correctly.
I am evaluating a treatment on a group of children. the class teachers completed the strengths and difficulties questionnaire on the same group of children at time 1(T1) and time 2 (T2)
When I initially interpreted the paired sample T-test, I was able to compare the means, and make assumptions as to whether the treatment produced an improvement or not.
For the Wilcoxon test, I have the medians. My difficulty is that for example, the hyperactivity scale
Scores from 0-5 are in the normal band, scores of 6 are borderline and scores that ranged from 7-10 are abnormal.
I’m finding it difficult to determine whether the conclusions I was able to make from the paired T-test will be the same for the Wilcox test. ( I did not check for normality, thus, now having to use the Wilcoxon test)
So for example,
For the hyperactivity scale for the paired t-test I found: (T1= time1 and T2 =time 2)
I then stated that: These results suggest that the treatment produced a significant improvement in the hyperactivity scale scores at T2. (I looked at the means for this, even though there was a decline, using the information above about the band scoring it suggested that there was an improvement) is this correct?
However, how do I make similar conclusions using medians?
For example the prosocial behaviour score band is as follows: 6-10 for normal, 5 =borderline and 0-4 is considered abnormal.
A test of normality reveals (see Appendix a) that the data was not normally distributed, except for the hyperactivity scale at T1 for the Kolmogorov-Smirnov test but not for the Shapiro-Wilk test. For this reason a non- parametric test was used.
A Wilcoxon T- test was carried out to compare the T1 scores to the T2 scores on the five SDQ scales and the total difficulties scores.
Table1 : Wilcoxon Signed Ranks for the five SDQ scales and the total difficulties scores at T1 and T2.
|
|
N |
Mean Rank |
Sum of Ranks |
Pro Social Scale at Time 1 - Pro Social Scale at Time 2 |
Negative Ranks |
37a |
30.38 |
1124.00 |
Positive Ranks |
19b |
24.84 |
472.00 | |
Ties |
19c |
|
| |
Total |
75 |
|
| |
Hyperactivity Scale at Time 1 - Hyperactivity Scale at Time 2 |
Negative Ranks |
5d |
21.50 |
107.50 |
Positive Ranks |
52e |
29.72 |
1545.50 | |
Ties |
18f |
|
| |
Total |
75 |
|
| |
Conduct Problems Scale at Time 1 - Conduct Problems Scale at Time 2 |
Negative Ranks |
12g |
19.83 |
238.00 |
Positive Ranks |
25h |
18.60 |
465.00 | |
Ties |
38i |
|
| |
Total |
75 |
|
| |
Emotional Symptoms Scale at Time 1 - Emotional Symptoms Scale at Time 2 |
Negative Ranks |
10j |
18.85 |
188.50 |
Positive Ranks |
34k |
23.57 |
801.50 | |
Ties |
31l |
|
| |
Total |
75 |
|
| |
Peer Problems Scale at Time 1 - Peer Problems Scale at Time 2 |
Negative Ranks |
12m |
26.58 |
319.00 |
Positive Ranks |
40n |
26.48 |
1059.00 | |
Ties |
23o |
|
| |
Total |
75 |
|
| |
Total SDQ score at Time 1 - Total SDQ score at Time 2 |
Negative Ranks |
8p |
34.12 |
273.00 |
Positive Ranks |
62q |
35.68 |
2212.00 | |
Ties |
5r |
|
| |
Total |
75 |
|
|
Table 2: Wilcoxon Signed Ranks Test
|
Pro Social Scale at Time 1 - Pro Social Scale at Time 2 |
Hyperactivity Scale at Time 1 - Hyperactivity Scale at Time 2 |
Conduct Problems Scale at Time 1 - Conduct Problems Scale at Time 2 |
Emotional Symptoms Scale at Time 1 - Emotional Symptoms Scale at Time 2 |
Peer Problems Scale at Time 1 - Peer Problems Scale at Time 2 |
Total SDQ score at Time 1 - Total SDQ score at Time 2 |
Z |
-2.685a |
-5.755b |
-1.735b |
-3.615b |
-3.467b |
-5.687b |
Asymp. Sig. (2-tailed) |
.007 |
.000 |
.083 |
.000 |
.001 |
.000 |
a. Based on negative ranks
b. Based on positive ranks
A Wilcox Signed Rank Test revealed that:
Am I correct in my interpretation here???
There was a significant difference in the pro-social behaviour scale scores for T1 and T2, Z=-2.685, p=0.07. The median score on the pro-social behaviour scale has decreased from T1 (Md=30.38) to T2 (Md=24.84). The results suggest that the treatment produced a significant difference in the pro-social behaviour scale scores. However, they were better at T1 than at T2.
There was a significant difference in the hyperactivity scale for T1 and T2, Z=-5.755, p<0.05. The median score on the hyperactivity scale has increased from T1 (Md=21.50) to T2 (Md=29.72). These results suggest that NA produced a significant improvement in the hyperactivity scale scores and that at T2.
Hope my questions make sense and that you are able to help!