Statistics for Data Analytics

PART A
The datafiles “OverseasTrips’ and ‘NewHouseRegistrations’, uploaded on Moodle have been sourced from
CSO.ie, the data repository of the Central Statistics Office in Ireland.
1. “OverseasTrips’ is a quarterly time series of overseas trips to Ireland by non-residents from Q1, 2012 to
Q4, 2019.
2. ‘NewHouseRegistrations’ is an annual series of new house registrations from 1978 to 2019.
You are required to estimate models for both time series. Your report should address the following aspects for
each series:
• An assessment of the components of the raw time series.

• Fit a minimum of three candidate time series models to each. You may wish to consider simple models,
exponential smoothing models and non-seasonal and/or seasonal ARIMA models.
• Perform and explain the appropriate diagnostic tests and checks and describe the rationale for your
choice of ‘optimum’ model for each series.

• Forecast each series for three periods ahead with appropriate prediction intervals and illustrative
graphics and provide your assessment of the adequacy of the chosen model.
PART B
The file ‘Childbirths’ contains data on childbirths in a US city.
Choose or create an appropriate dichotomous variable from the data and estimate and report on a suitable
binary logistic regression model to classify the cases in respect of the chosen dependent variable.
Use PCA or
an alternative dimension reduction technique if you consider it necessary.
Your report should incorporate:
• Descriptive statistics and appropriate visualisations to enhance understanding of the variables used.
• Details of the model building steps you undertook in the process of arriving at your final model.
Reasons
for rejection of intermediate models should be explained clearly.
• Appropriate diagnostics and an assessment of model fit.
• A succinct summary of the parameters of your final model explaining the inclusion and influence of the
independent variables.