Bao-Tran Nguyen-Phuoc z5167599

PSYC3001 ASSIGNMENT 2
Part 1
A) i. Rationale
A study investigated the effects of varying non-psychological and psychological
treatments for a population of high school students diagnosed with IAD (Internet
Addiction Disorder). Fifty-five teenagers were randomly allocated to one of five
different conditions (J = 5, n = 11, N = 55):
Treatment 1 = Pharmacological intervention – SSRI (antidepressant; Cash, D.
Rae, H. Steel & Winkler, 2012)
Treatment 2 = Positive psychology intervention (group therapy designed to
improve quality of social relationships; Khazaei, Khazaei &
Ghanbari-H., 2017)
Treatment 3 = Physical exercise intervention (Pan, 2020)
Treatment 4 = Psychological intervention (combination of group therapy and
individual therapy using cognitive behavioural therapy (CBT),
games and systemic desensitisation)
Treatment 5 = Comprehensive therapy (combination of Physical exercise
intervention with Psychological intervention; Pan, 2020)

The dependent variable was the difference in daily internet usage before and after 6-
weeks of treatment, operationalised as hours spent online after treatment – hours
spent online before treatment (such that a negative score indicated change in daily
internet usage in hours). The smallest difference regarded to be of clinical
importance would be a difference of 3 hours.

ii. Table of means:

Groups 1: Pharmacological 2: Positive 3: Physical 4: Psychological 5: Comprehensive
psychology exercise intervention therapy
Means (Mj) -5 -4 -6.5 -3.3 -10
Bao-Tran Nguyen-Phuoc z5167599

B) Overall test
i. ANOVA summary table

ii. Decision rule: Reject H0: μ1 = μ2 = μ3 = … = μj (homogeneity hypothesis) if F >

Fc
Outcome: Fc= Fα;ν1 ,ν2 = F0.05;4,50 = 2.56; F = 21.292
F > Fc ; Reject H0: μ1 = μ2 = μ3 = … = μj (homogeneity hypothesis)
Conclusion: For a population of students, population means are
heterogenous (i.e. effect parameters are not all zero).
C) i.
Contrasts Rationale
Ψ1 = μ1 – μ3 Comparing the difference in average daily internet usage between
those who took the pharmacological treatment and those who had
physical exercise. This contrast looks at the difference between the two
non-psychological treatments (related to Ψ3)
Ψ2 = μ2 – μ4 Comparing the difference between positive psychology and
psychological intervention. This contrast looks at the difference
between the two psychological treatments (related to Ψ3)
( μ 1+ μ 3) Comparing the difference between the non-psychological treatments
Ψ3 = -
2
(Pharmacological and Physical exercise) and the psychological
( μ 2+ μ 4)
treatments (Positive psychology and psychological).
2
μ 1+ μ 2+ μ 3+ μ 4 Comparing the difference between the single-therapy treatments
Ψ4 =
4
(Pharmacological + Positive psychology + Physical exercise +
– μ5
Psychological) and combination treatment (Comprehensive). This
contrast will look at whether combination treatment has more of an
effect on reducing internet usage compared to single-therapy
treatments.
Bao-Tran Nguyen-Phuoc z5167599

( μ 2+ μ 4) Comparing the difference between the psychological therapies (Positive

Ψ5 = – μ5
2
psychology + psychological) and Comprehensive treatment. This
contrast was suggested by the coefficients of the maximal contrast.

Contrasts of interest: [ 0 1 0 1 -2 ]

ii. Table of contrast coefficients

Contrasts Mean difference Integer {m, r}
coefficients coefficients
Ψ1 = μ1 – μ3 [ 1 0 -1 0 0 ] [ 1 0 -1 0 0 ] {1, 1}
Ψ2 = μ2 – μ4 [ 0 1 0 -1 0 ] [ 0 1 0 -1 0 ] {1, 1}
( μ 1+ μ 3) ( μ 2+ μ 4) 1 1 1 1 [ 1 -1 1 -1 0 ] {2, 2}
Ψ3 = - [ - - 0]
2 2 2 2 2 2
μ 1+ μ 2+ μ 3+ μ 4 1 1 1 1 [ 1 1 1 1 -4 ] {4, 1}
Ψ4 = – [ -1 ]
4 4 4 4 4
μ5
( μ 2+ μ 4) 1 1 [ 0 1 0 1 -2 ] {2, 1}
Ψ5 = – μ5 [0 0 -1 ]
2 2 2

D) Contrasts Analysis (Scheffé)

i. Test and CI output from PSY
Contrast Fc = 1 × F; 1, 2 = 10.24 (*) indicates significant test outcome

Analysis of Variance Summary Table

Source SS df MS F
------------------------------------------------
Between
------------------------------------------------
B1 12.375 1 12.375 3.115
B2 7.564 1 7.564 1.904
B3 60.045 1 60.045 15.115 *
B4 258.339 1 258.339 65.033 *
B5 318.121 1 318.121 80.082 *
Error 198.622 50 3.972
------------------------------------------------
 Bao-Tran Nguyen-Phuoc z5167599

Raw CIs (scaled in Dependent Variable units)

-------------------------------------------------------
Contrast Value SE ..CI limits..
Lower Upper
-------------------------------------------------------
B1 1.500 0.850 -1.218 4.218
B2 -1.173 0.850 -3.891 1.545
B3 -2.336 0.601 -4.258 -0.414
B4 5.418 0.672 3.269 7.567
B5 6.586 0.736 4.232 8.940
-------------------------------------------------------

iii. Decision rule: If SS  g > SSc; Reject H0: Ψg = 0

F; 1, 2 = 2.56; 1 = J – 1 = 4; 2 = N – J = 50
SSc = 1 × F; 1, 2 × MSE = 4 x 2.56 x 3.972 = 40.67

Contrasts Outcomes
Ψ1 = μ1 – μ3 Ψ = -5 – (-6.5) = 1.5
Σc2 = 12 + (-1)2 = 2
2
n(Ψ ) 11(1.5)2
SS 1 = = = 12.375
Σ c2 2
SS 1 < SSc ; Fail to reject H0.
Ψ2 = μ2 – μ4 Ψ = -0.7
Σc2 = 12 + (-1)2 = 2
2
n(Ψ ) 11(−0.7)2
SS 1 = = = 2.695
Σ c2 2
SS 1 < SSc ; Fail to reject H0.
( μ 1+ μ 3) ( μ 2+ μ 4) Ψ = [(1/2)M1 + (1/2)M3] – [(1/2)M2 + (1/2)M4]
Ψ3 = -
2 2
= [(1/2)*-5 + (-1/2)*-6.5] – ([1/2)*-4 + (-1/2)*-3.3]
= 2.5 + 3.25 – (2 + 1.65) = 2.1
Σc2 = (.05)2 + (-0.5)2 + (-0.5)2 + (0.5)2 = 1
2
n(Ψ ) 11(2.1)2
SS 1 = 2 = = 48.51
Σc 1
SS 1 > SSc ; Reject H0.
μ 1+ μ 2+ μ 3+ μ 4 Ψ = 5.3
Ψ4 = – μ5
4
Σc2 = .252 + .252 + .252 + .252 + -12 = 1.25
 Bao-Tran Nguyen-Phuoc z5167599

2
n(Ψ ) 11(−14.7)2
SS 1 = 2 = = 1901.59
Σc 1.25
SS 1 > SSc ; Reject H0.
( μ 2+ μ 4) Ψ = -13.65
Ψ5 = – μ5
2
Σc2 = .52 + .52 + 12 = 1.5
2
n(Ψ ) 11(−13.65)2
SS 1 = = = 1366.365
Σ c2 1.5
SS 1 > SSc ; Reject H0.

E) Interpretation of contrasts analysis

The raw CI for B4 suggest that the difference between single therapy treatments
and comprehensive treatment is positive and clinically important. On average,
Treatments 1-4 reduce internet usage by as much as 7.56 hours or at least 3.269
hours compared to the comprehensive treatment, however it is not clear whether
there are any clinically important differences amongst these treatments (B1, B2).

B3 suggests an important difference favouring psychological treatments over non-

psychological treatments by as much as 4.258 hours, but this difference could
possibly be trivial (upper limit is close to 0).

B5 suggests a positive effect of clinical importance where psychological treatments

can reduce internet usage more than the Comprehensive treatment by at least 4.232
hours or as much as 8.94 hours. B5 also contributes to the inference for B4 as it
accounts for some of the treatment effect (in favour of Treatments 2+4 over
Treatment 5).

No directional inferences can be made for B1 and B2 as there is insufficient

precision in their raw CIs (both CIs contain both positive and negative values, either
of which may be of clinical importance and contain 0).
Bao-Tran Nguyen-Phuoc z5167599

Part 2 – Planned contrast (Bonferroni)

Fc = 7.171 (k = 5); CC = 2.7 (whereas Scheffe CC = 3.2, Fc = 4 x 2.56 = 10.24).
Scheffe Fc > Bonferroni Fc

Raw CIs (scaled in Dependent Variable units)

-------------------------------------------------------
Contrast Value SE ..CI limits..
Lower Upper
-------------------------------------------------------
B1 1.500 0.850 -0.776 3.776
B2 -1.173 0.850 -3.448 1.103
B3 -2.336 0.601 -3.946 -0.727
B4 5.418 0.672 3.619 7.217
B5 6.586 0.736 4.616 8.557
-------------------------------------------------------

Do the inferences from this output differ from inferences made in Part 1?
No. Although CC for Bonferroni method is smaller than Scheffe CC (and therefore
produces narrower, more precise CIs), the planned contrasts analysis does not change
inferences made from Part 1. B1 and B2 remain non-significant and B3-5 are still
significant with similar directional inferences.

We know that for planned contrasts, Bonferroni method will always produce a
smaller contrast Fc value and thus give more precise SCIs (given that k ≤ J – 1). When k
> J – 1, this is not the case as Fc increases as a function of k, in which case Scheffe Fc
will eventually be less than Bonferroni Fc because it is not dependent on the number
of planned contrasts (k) and thus the Fc value for Scheffe remains the same regardless
of k. However, for this set of contrasts Bonferroni Fc < Scheffe because k is not that
much larger relative to J – 1 and will have more statistical power (a smaller Fc makes it
easier to reject the null hypothesis).
 Bao-Tran Nguyen-Phuoc z5167599

Part 3 – Planned analysis controlling for PCER (Orthogonal contrasts)

Contrasts Integer coefficients

Ψ1 = μ1 – μ3 [ 1 0 -1 0 0 ]

Ψ2 = μ2 – μ4 [ 0 1 0 -1 0 ]
( μ 1+ μ 3) ( μ 2+ μ 4) [ 1 -1 1 -1 0 ]
Ψ3 = -
2 2
μ 1+ μ 2+ μ 3+ μ 4 [ 1 1 1 1 -4 ]
Ψ4 = 4
– μ5

Raw CIs (scaled in Dependent Variable units)

-------------------------------------------------------
Contrast Value SE ..CI limits..
Lower Upper
-------------------------------------------------------
B1 1.500 0.850 -0.207 3.207
B2 -1.173 0.850 -2.880 0.534
B3 -2.336 0.601 -3.543 -1.129
B4 5.418 0.672 4.069 6.768
-------------------------------------------------------

Groups 1: Pharmacological 2: Positive 3: Physical 4: Psychological 5: Comprehensive

psychology exercise intervention therapy
Means (Mj) -5 -4 -6.5 -3.3 -10

How satisfactory is this analysis in providing an account of the data?

Not very satisfactory. B4 tells us that the single therapy treatments are more effective
than the comprehensive treatment, but given the fact that the Comprehensive treatment is
a combination of 2 of the single therapy treatments (Physical exercise and Psychological
intervention) two of the contrasts in the set should compare the effectiveness of each
individual therapy against the combination of the two (i.e. Physical exercise vs.
Comprehensive, and Psychological intervention vs Comprehensive treatment) in order to
Bao-Tran Nguyen-Phuoc z5167599

have a more complete account of the data. Additionally, the pattern of the means suggest
that the average of Treatments 1-4 is less than the mean for Comprehensive therapy
whereas the analysis suggests the opposite.

Furthermore, B1 and B2 are not significant and so no inference can be made about the
difference between different types of non-psychological/psychological treatments. For
example, Pharmacological treatment may be more effective than Physical exercise but there
may be also no difference. This does not account for the pattern of means (Pharmacological
mean difference is less than Physical exercise).

B3 did provide a significant outcome and suggests that psychological treatments are
more effective than non-psychological treatments. However, it would also be helpful to see
the difference between the different psychological treatments (B2; Positive psychology vs.
psychological intervention), for example which psychological therapy contributed more to
the treatment effect shown in B3 or do they have similar effects? The analysis output from
B3 also does not follow the pattern of means (mean values for non-psychological treatments
> psychological treatments).

In summary, due to lack of significant outcomes and the incoherence of the analysis
output to pattern of means, the orthogonal analysis does not provide a satisfactory account
of the data.

References:
Cash, H., D. Rae, C., H. Steel, A., & Winkler, A. (2012). Internet Addiction: A Brief
Summary of Research and Practice. Current Psychiatry Reviews, 8(4), 292-298. doi:
10.2174/157340012803520513
Khazaei, F., Khazaei, O., & Ghanbari-H., B. (2017). Positive psychology interventions for
internet addiction treatment. Computers In Human Behavior, 72, 304-311. doi:
10.1016/j.chb.2017.02.065
Pan, T. (2020). PSYCHOLOGICAL AND EXERCISE INTERVENTIONS FOR TEENAGERS WITH
INTERNET ADDICTION DISORDER. REVISTA ARGENTINA DE CLINICA PSICOLOGICA, 29(2), 226-
231. doi: 10.24205/03276716.2020.228
Bao-Tran Nguyen-Phuoc z5167599