Survival Analysis and Hypothesis Testing: Statistical Test Selection

This assignment solution focuses on survival analysis and hypothesis testing across eight different scenarios. Each scenario presents a null and alternative hypothesis related to health outcomes and patient well-being, such as HRQoL in CHB patients, condom usage among adolescents, alcohol-related hospitalizations, effectiveness of antenatal care, mortality rates, the impact of social networks on depression, the effectiveness of mental health programs, and the efficacy of a new glucose monitoring instrument. The solution identifies appropriate statistical tests for each hypothesis, including Cox regression, logistic regression, Chi-squared tests, t-tests, ANOVA, Kruskal-Wallis test, and the long-rank test, providing justifications for the chosen tests based on the nature of the data and research questions. The document demonstrates the application of statistical methods to analyze survival data and draw meaningful conclusions about various health-related scenarios, making it a valuable resource for students studying statistics and biostatistics.

Survival-Analysis and Censoring

Abbreviations:
HRQoL – Health related quality of life
CHB – Chronic Hepatitis B
Scenario 1
Hypothesis:
Null hypothesis – The HRQoL score of CHB patients is higher than expected and may worse
when disease become active
Alternative Hypothesis – The HRQoL score of CHB patients is lower than expected and may
worse when disease become active
Hypothesis test: Cox Regression Model, with t-test and ANOVA analysis
Justification of test: Since 300 patients are taken for conducing a survey to evaluate the
HRQoL of them by dividing them into three clinical group, so applying regression model helps
in investigating the contribution of explanatory factors on risk of events.
Scenario 2
Hypothesis:
Null hypothesis – Adolescents with known to be HIV negative expected to use condoms as
compared to those with unknown HIV status
Alternative Hypothesis – Adolescents with known to be HIV negative are not expected use
condoms as compared to those with unknown HIV status
Hypothesis test: Logistic regression and Chi-squared test
Justification of test: As survey is conducted by dividing the respondents into two main groups,
therefore binary logistic regression models helps in comparing the two condom user groups at
same time, while chi-squared test is used to investigate the difference between these two groups
for testing hypothesis
1

Scenario 3
Hypothesis:
Null hypothesis – Alcohol-related hospitalisation rates in Local Authority areas in England are
considerably higher for males than females
Alternative Hypothesis – Alcohol-related hospitalisation rates in Local Authority areas in
England are considerably not higher for males than females
Hypothesis test: t-test
Justification of test: As hospital data is right skewed i.e. normally distributed therefore, t-test
or test statistics helps in identifying the significant difference between groups.
2

Scenario 4
Hypothesis:
Null hypothesis – Antenatal Care proves effective in reducing the risk of maternal and newborn
mortality, intrauterine growth restrictions, prematurity and stillbirth
Alternative Hypothesis – Antenatal Care doesn't prove effective in reducing the risk of maternal
and newborn mortality, intrauterine growth restrictions, prematurity and stillbirth
Hypothesis test: Chi-squared test and Logistic regression analysis
Justification of test: Logistic regression analysis is mainly used when dependent variable in
study is binary, which helps in determining the effects of antenatal care
In the same study researchers wanted to find out if the mortality rate was higher in babies of
mothers who attended antenatal clinics in rural Uganda (n=284) compared to the
mortality rate of babies born to mothers who did not attend antenatal clinics (n=298).
Scenario 5
Hypothesis:
Null hypothesis – The mortality rate was higher in babies of those mothers of rural Uganda who
have attended the antenatal clinics
Alternative Hypothesis – The mortality rate was not higher in babies of those mothers of rural
Uganda who have attended the antenatal clinics
Hypothesis test: two-talied test of ANOVA with 95% confidence interval
Justification of test: As depended variable is categorical and has two variables, therefore,
ANOVA test is used to test the hypothesis
3

Scenario 6
Hypothesis:
Null hypothesis – Older people who are having poor or limited social networks are more likely
to face depression
Alternative Hypothesis – Older people who are having poor or limited social networks are not
likely to face depression
Hypothesis test: Chi-squared test
Justification of test: As data is categorised on demographic based and normally distributed, as
well as divided in three groups therefore, Chi-squared test is applied
Scenario 7
Hypothesis:
Null hypothesis – People who attend the mental health programme in rural Kenya are more
likely to assess their well-being state within 6 months
Alternative Hypothesis – People who attend the mental health programme in rural Kenya are
not more likely to assess their well-being state within 6 months
Hypothesis test: Kruskal-Wallis test
Justification of test: As data is not normally distributed therefore, non-parametric test like
Kruskal-Wallis test can be applied on two different or equal sample size
4

Scenario 8
Hypothesis:
Null hypothesis – New instrument proves effective for self-testing and monitoring the glucose
in people with type 2 diabetes
Alternative Hypothesis – New instrument doesn't prove effective for self-testing and monitoring
the glucose in people with type 2 diabetes
Hypothesis test: The long-rank test
Justification of test: As to test the effectiveness of new instrument for self-testing glucose
required all available data, therefore, long-rank test is used.
5