You are on page 1of 9

Name : Danik Iga Prasiska (Indonesia)

2nd Batch Global Health Security

1. Complete demographics regarding numerical outcomes to see there are differences between three diet
methods. As a post-hoc analysis, suppose we are only interested in diet 1.

Answer
Because in the question, there is no note that whilst controlling for scale covariates. Therefore, the analysis
will be performed as One Way ANOVA.

a. Assumption

Independent
We assume that the three set of data collected constitute independent simple random from three indicated
populations.

Data Normality

Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Statistic df Sig. Statistic df Sig.
Age .060 78 .200* .987 78 .645
Height .133 78 .002 .943 78 .002
*
Weight before the diet .062 78 .200 .969 78 .055
Weight after 10 weeks on
.085 78 .200* .958 78 .011
the diet
weightLOST .057 78 .200* .990 78 .802
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction

From the table Test of Normality we can see that the significant level (p-value) of several variable in Saphiro –
Wilk test > 0.05. It indicates if the data distribution is normal. Except, for Height and Weight after 10 weeks
on the diet because the significant level (p-value) of these data are < 0.05.
Therefore, we should look at the Q-Q plots to assume the normality of the data for height and weight after 10
weeks on the diet.

From the Q-Q plots of Height, we can see that the data follow the line. However, there is skewness in the end of
the lower and upper line. Since the majority of data follow the line, we could assume that the data is close to
normally distributed.

1
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

From the Q-Q plots of Weight after 10 weeks, we can see that the data follow the line. However, there is skewness
in the end of the lower and upper line. Since the majority of data follow the line, we could assume that the data is
close to normally distributed.

Equal Variance

Test of Homogeneity of Variances


Levene Statistic df1 df2 Sig.
Age .142 2 75 .868
Height .795 2 75 .455
Weight before the diet 1.177 2 75 .314 All of sig. (p-value) > 0.05

Weight after 10 weeks on


.433 2 75 .650
the diet
weightLOST .659 2 75 .520

From the table of Homogeneity of Variances, we could see that the significant level (p-value) of all the data >
0.05. It is indicated that the variance of all the data are equal.

Conclusion of Assumption:

We assume that all the data are independent to each other, normally distributed with equal variance. Therefore,
the data meet the criteria to perform parametric test, in this case One-Way ANOVA.

2
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

b. Hypothesis

Age H0 = On average there is no difference of age in three diets group


Ha = At least there is one difference of age mean between three diets group

Height H0 = On average there is no difference of height in three diets group


Ha = At least there is one difference of height mean between three diets group

Pre weight H0 = On average there is no difference of pre weight in three diets group
Ha = At least there is one difference of pre weight mean between three diets group

Weight 10week H0 = On average there is no difference of weight 10 week in three diets group
Ha = At least there is one difference of weight 10 week mean between three diets group

Weight Lost H0 = On average there is no difference of weight lost in three diets group
Ha = At least there is one difference of weight lost mean between three diets group

α=5%

c. Test Statistic

Post-hoc p-values
adjusted by Bonferroni

Total Diet 1 Diet 2 Diet 3 p-values Diet 2 vs 1 Diet 3 vs 1

Age 39.15±9.81 40.88±9.72 39.00±9.51 37.78±10.32 0.5350 >0.9999 0.5340

Height 170.82±11.28 170.29±10.95 174.85±12.08 167.26±9.71 0.0430 0.2840 0.6540

Pre weight 72.53±8.72 72.88±8.38 71.11±10.09 73.63±7.61 0.5600 0.9520 >0.9999

Weight 10
68.68±8.92 69,58±8.40 68.09±10.22 68.48±8.42 0.8330 >0.9999 >0.9999
weeks
Weight
3.84±2.56 3.30±2.24 3.03±2.52 5,15±2.40 0.0030 >0.9999 0.0140
LOST

From the table, we can make conclusion that:

For Age variable, the oldest age mean is in the Diet 1 group with 40.88 kg and standard deviation ±9.72.
The p-value is (0.5350) > 0.05. Therefore, it is failed to reject H0. It can be concluded that data of age
does not have enough evidence to prove that there is at least 1 difference between group (There is no
difference of age mean in three diets group)

For Height variable, the highest height mean is in the Diet 2 group with 174.85 cm and standard
deviation ±12.08. The p-value is (0.0430) < 0.05. Therefore, it is reject H0. So, there is at least 1
difference between three group. However, when we perform Post-hoc test comparing Diet 2 with 1 and
Diet 3 with 1, both of the p-value are > 0.05. So the difference is not in group diet 2 with diet 1 or diet 3
with diet 1. So, the difference may occur in group Diet 2 compare to Diet 3 group.

For pre weight variable, the heaviest pre weight mean is in the Diet 3 group with 73.63kg and standard
deviation ±7.61. The p-value (0.0560) > 0.05. Therefore, it is failed to reject H0. It can be concluded
that data of age does not have enough evidence to prove that there is at least 1 difference between group
(There is no difference of pre weight mean in three diet groups.

3
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

For weight after 10weeks variable, the heaviest weight after 10weeks mean is in the Diet 1 group with
69.58 kg and standard deviation ±7.61. The p-value (0.0560) > 0.05. Therefore, it is failed to reject H0.
It can be concluded that data of age does not have enough evidence to prove that there is at least 1
difference between group (There is no difference of weight after 10weeks mean in three diet groups)

For weight Lost variable, the most weight lost mean is in the Diet 3 group with 5.15 kg and standard
deviation 2.40. The p-value is (0.0030) < 0.05. Therefore, it is reject H0. So, there is at least 1 difference
of weight lost between three group. After perform post hoc test, it is known that the p-value for Diet 3
compare to Diet 1 is (0.0140) < 0.05. So, the difference is between Diet 3 and Diet 1 group.

Mean Plot
a. Age Variable
From the mean plot of age variable, it is
clearly seen that the oldest mean is in Diet
1 group and the youngest mean is in the
Diet 3 group.

b. Height Variable
From the mean plot of Height variable, it
is clearly seen that the highest mean of
height is in the Diet 2 group and the
shortest height mean is in the Diet 3
group.

4
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

c. Pre Weight Variable


From the pre weight variable mean plot, it
can be seen that the heaviest pre weight
mean is in the Diet 3 group and the most
lightweight is in Diet 2 group.

d. Weight after 10weeks Variable


Mean plot of weight after 10 weeks
variable, showed that the heaviest weight
after 10 weeks mean is in the Diet 1 group
and the lowest in Diet 2 group.

e. Weight Lost Variable


Mean plot of weight lost variable showed
that the most weight lost is in the Diet 3
group and the least is in the Diet 2 group.

5
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

2. Use the multiple regression to predict the hospital aPTT from the CoaguCheck aPTT level as well as
the medication received. Is knowledge of medication useful in the prediction? Let 𝜶 =0.05 for all tests.
Set “Warfarin and Enoxaparin” group as reference group.

Answer
a. Assumption

Independent
We assume that the Y data (partial thromboplastin time using standard laboratory hospital assay (hospital
aPTT)) is independent.

Relationship among variables coaguCheck a PTT with Hospital aPTT

Scatter matrix plot above, indicates that there is linear association between coaguCheck aPTT with
Hospital aPTT.

Data distribution

From the plot provided, we can assume that the data distribution for hospital a PTT is close to normal or normally
distributed because the point to cluster around the horizontal line.

6
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

Equal Variance

The scatter plot of residual provided above, showed that there is no pattern formed by the point. It indicates
weather that the relationship of the predictor outcome is linear and weather that equal variance assumption is
met.

Conclusion of assumption: the data meet all of the assumption criteria which are dependent variable are
independent, linear association between independent and dependent variable, dependent variable normally
distributed, and equal variance.

b. Hypothesis

CoaguCheck aPTT H0 : β1 = 0 α = 5%
Ha : β ≠ 0

Heparin H0 : β2 = 0
Ha : β2 ≠ 0

Heaprin & Warfarin H0 : β3 = 0


Ha : β3 ≠ 0

7
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

c. Test Statistic
Correlations
hosp coagu Heparin HeparWar
hosp 1.000 .735 .366 .151
coagu .735 1.000 .173 .090
Pearson Correlation
Heparin .366 .173 1.000 -.500
HeparWar .151 .090 -.500 1.000
hosp . .000 .000 .078
coagu .000 . .052 .199
Sig. (1-tailed)
Heparin .000 .052 . .000
HeparWar .078 .199 .000 .
hosp 90 90 90 90
coagu 90 90 90 90
N
Heparin 90 90 90 90
HeparWar 90 90 90 90

From the correlation table, it can be seen that the correlation between hospital aPTT and CoaguCheck aPTT is,
0.735, which means higher coaguCheck aPTT level tends to have higher hospital aPTT level. The correlation
level of heparin treatment and heparin & warfarin are 0.366 and 0.151, which mean that get heparin or heparin &
warfarin medication tend to have higher hospital aPTT level.

Model Summaryb
Model R R Square Adjusted R Std. Error of the
Square Estimate
a
1 .813 .661 .649 15.49515
a. Predictors: (Constant), HeparWar, coagu, Heparin
b. Dependent Variable: hosp

From the model summary table, it showed that the R square is 0.649, meaning that the proportion of variance in
the dependent variable that can be explained by independent variable is 64,9%. It is indicated that the model fit
well.

ANOVAa
Model Sum of Squares df Mean Square F Sig.
Regression 40310.465 3 13436.822 55.964 .000b
1 Residual 20648.560 86 240.100
Total 60959.025 89
a. Dependent Variable: hosp
b. Predictors: (Constant), HeparWar, coagu, Heparin

ANOVA Table showed that the overall p-value is < 0.0001 which means this model is statistically significant.

8
Name : Danik Iga Prasiska (Indonesia)
2nd Batch Global Health Security

Coefficientsa
Model Unstandardized Standardized t Sig. 95.0% Confidence Interval
Coefficients Coefficients for B
B Std. Error Beta Lower Bound Upper Bound
(Constant) 15.457 3.871 3.993 .000 7.761 23.152
coagu .558 .057 .639 9.808 .000 .445 .671
1
Heparin 22.229 4.135 .403 5.376 .000 14.009 30.450
HeparWar 16.252 4.090 .294 3.974 .000 8.122 24.381
a. Dependent Variable: hosp

The significant level of CoaguCheck, Heparin, and Hearin&Warfarin variable are <0.0001 which are less than
0.05. It is indicated that H0 for CoaguCheck aPTT, Heparin, and Hearin&Warfarin variable are rejected.

From the coefficients table, it can be estimated that the regression model is:

yj = β0 + β1X1j + β2X2j + ej

yj = 15.475 + 0.558*coaguj + 22.229 *I[Heparin]j + 16.252*I[Heparin&Warfarin]j

From the model, we can make conclusion that CoaguCheck aPTT and medication are useful to predict hospital
aPTT level. It predicts that every time we increase CoaguCheck aPTT level by 1, it will increase the hospital
aPTT level by 0.558.
The I[Heparin]j denotes that the indicator function and it has value 1 when get medication heparin, and 0 if not,
otherwise. If the patient gets medication Heparin only, the Hospital aPTT level will increase by 1 *22.229 =
22.229, while if the patient gets medication Heparin&Warfarin, it will increase Hospital aPTT level by 1*16.252
= 16.262.
Therefore, this model can answer the question, the knowledge of medication is useful in hospital aPTT level
prediction.