The markers will mark based only on correct output and comments.

Questions 1 and 2 use data from the New Zealand Quality in Healthcare Study, which assessed the occurrence,     impact and preventability of adverse events recorded in New Zealand public hospitals in 1998. Documentation can

be found in New Zealand Quality of Healthcare Study.pdf

Read the data set in NZqhs.240.csv into iNZight Lite.

First, specify a survey design with:

wt as the weighting variable,

stratum_id as the strata variable, and

cluster_id as the 1st stage clustering variable.

1.   Associations                                                                                                                                                                   7 MARKS (a) Produce the summary for the stay variable – Length of hospital stay – by sex – male and female.  (1 MARKS).

(b) Use a t-test to test whether the length of hospital stay differs for males and females. Copy the output into     your assignment.  What do you conclude? (you need to comment about significance and, if significant, direction

and size of effect)                                                                                                                                                             (2 MARKS)

(c) Create a categorical variable for mdcgp – major diagnostic category.

NOTE: turn off survey design before creating this variable, then specify the survey design again.

Rename the levels according to the levels listed in New Zealand Quality of Healthcare Study.pdf, and name the final variable mdcgp_cat.  Produce the summary for the stay by mdcgp_cat.  Which major diagnostic category

has the longest mean length of stay? Which has the shortest?                                                                               (2 MARKS)

(d) Use an ANOVA to test whether length of stay differs by major diagnostic category.  Copy the output into

your assignment.  Is there evidence that length of stay differs by major diagnostic category?                   (2 MARKS)

2.    Graphs                                                                                                                                                                                              7 MARKS

Convert aey – Adverse event –to a categorical variables.

NOTE: turn off survey design before converting this variable, then specify the survey design again.

Rename the levels according to the levels listed in New Zealand Quality of Healthcare Study.pdf, and name the final variable aey_cat.

(a) Create a bar graph of your aey_cat (variable 1) – by mdcgrp_cat (variable 2).                                       (2 MARKS)

From the graph answer:

(b) Which major diagnostic category is most likely to result in an adverse event?  Which is least likely?

(2 MARK)

(c) Conduct a chi-squared test of the association between aey_cat and mdcgrp_cat.  Is there evidence for a difference in the likelihood of an adverse event between major diagnostic categories?

(2 MARKS)

(d) What is the estimated difference in proportion with an adverse event between musculoskeletal admissions   and neonatal admissions?                                                                                                                                       (1 MARKS)

3.     Stratified sampling                                                                                                                                                                     6 MARKS

The table below shows the population sizes for each of New Zealand’s 16 regions, as at the 2018 Census.

Suppose a survey sampled 1000 people from each region.

(a)   Calculate the sampling fractions for Auckland and West Coast to five decimal places.     (b)  What weight should Auckland and West Coast be given in analyses (one decimal place)?

(c)  What is the total sample size?

(d)  What is the sum of weights across all people in the sample?