讲解MAS6019、讲解MAS6061留学生、辅导Python设计、辅导Java/C/C++程序 调试Matlab程序|讲解R语言编程
- 首页 >> Web MAS6019 / MAS6061 Time Series Project
Analysis of Daily Temperatures in Melbourne
A time series data set consisting of daily maximum temperatures (0C)
in Melbourne can be found in the file TempMelb1
. The data set covers a
period of 1 January 1981 to 31 December 1990 and it is kindly provided
for educational use by the Time Series Data Library and the data provider
DataMarket ( DataMarket.com ).
Write a report on the data, concentrating on (a) describing their structure
and (b) discussing time series modelling and forecasting. The report will
account for 15% of the Semester 2 assessment of this module. For the deadline
of submission, consult the Course work schedule.
Some comments/suggestions and notes on organisational matters follow
(however, note that they are not mandatory). There is no page limit for the
report.
Comments and suggestions
1. A Box-Jenkins approach to part (a) of the question would use differencing
to ensure stationarity. Model diagnostics can be used to ensure
that the model fit is acceptable. However, it has to be recognised that
differencing has limitations. You will find that differencing for a very
long seasonal cycle does not work in R. You will have to deal with this
issue in the project.
2. In the light of (1), a direct approach of modelling, which does not need
to rely on the differencing process, could be the use of an appropriate
state space model.
3. To address forecasting you may decide to provide forecasts for timepoints
in the end of the series. Another approach could be to split the
data in parts and forecast already known values pretending they were
not available to you initially.
Notes on Procedures and your Report
1. This is an assessed piece of work, so answers to questions about it
must be available to everyone equally. Any questions should therefore
be posted on the discussion board.
1On the course web page MAS6019 Semester 2 MSc Project MOLE web page2. Distance Learners have not yet had the benefit of participating in
MAS6005, so the criteria that are laid down for reports in MAS6005
will not be applied here. In particular, there will be no explicit consideration
of presentation issues in the assessment. Of course, what you
present has to be intelligible, otherwise I cannot mark it.
3. There is no page limit, but you need to use the space wisely. Very short
reports are not likely to cover the ground (especially with a few plots)
and very long reports are likely to be repetitive.
4. The body of the report should be in connected English, illustrated if
appropriate by suitable plots (though plots should appear only if they
are relevant to the argument and only if they are referred to explicitly
in the body of the report). The main body of the report should not
contain non-graphical software (e.g. R) output or jargon; you may put
annotated software output in appendices if you think it is important to
have it on record.
5. You should write the report as thought for an intelligent and statisticallytrained
reader (another MSc student for example) who knows the general
technical background of time series, but has not met these data
before, nor the software you use.
6. The report should be self-contained; it should not call for calculations
or clairvoyance on the part of the reader.
7. The report should be written so that the reader does not need to look
at appendices unless he/she wants to check something you have done.
Analysis of Daily Temperatures in Melbourne
A time series data set consisting of daily maximum temperatures (0C)
in Melbourne can be found in the file TempMelb1
. The data set covers a
period of 1 January 1981 to 31 December 1990 and it is kindly provided
for educational use by the Time Series Data Library and the data provider
DataMarket ( DataMarket.com ).
Write a report on the data, concentrating on (a) describing their structure
and (b) discussing time series modelling and forecasting. The report will
account for 15% of the Semester 2 assessment of this module. For the deadline
of submission, consult the Course work schedule.
Some comments/suggestions and notes on organisational matters follow
(however, note that they are not mandatory). There is no page limit for the
report.
Comments and suggestions
1. A Box-Jenkins approach to part (a) of the question would use differencing
to ensure stationarity. Model diagnostics can be used to ensure
that the model fit is acceptable. However, it has to be recognised that
differencing has limitations. You will find that differencing for a very
long seasonal cycle does not work in R. You will have to deal with this
issue in the project.
2. In the light of (1), a direct approach of modelling, which does not need
to rely on the differencing process, could be the use of an appropriate
state space model.
3. To address forecasting you may decide to provide forecasts for timepoints
in the end of the series. Another approach could be to split the
data in parts and forecast already known values pretending they were
not available to you initially.
Notes on Procedures and your Report
1. This is an assessed piece of work, so answers to questions about it
must be available to everyone equally. Any questions should therefore
be posted on the discussion board.
1On the course web page MAS6019 Semester 2 MSc Project MOLE web page2. Distance Learners have not yet had the benefit of participating in
MAS6005, so the criteria that are laid down for reports in MAS6005
will not be applied here. In particular, there will be no explicit consideration
of presentation issues in the assessment. Of course, what you
present has to be intelligible, otherwise I cannot mark it.
3. There is no page limit, but you need to use the space wisely. Very short
reports are not likely to cover the ground (especially with a few plots)
and very long reports are likely to be repetitive.
4. The body of the report should be in connected English, illustrated if
appropriate by suitable plots (though plots should appear only if they
are relevant to the argument and only if they are referred to explicitly
in the body of the report). The main body of the report should not
contain non-graphical software (e.g. R) output or jargon; you may put
annotated software output in appendices if you think it is important to
have it on record.
5. You should write the report as thought for an intelligent and statisticallytrained
reader (another MSc student for example) who knows the general
technical background of time series, but has not met these data
before, nor the software you use.
6. The report should be self-contained; it should not call for calculations
or clairvoyance on the part of the reader.
7. The report should be written so that the reader does not need to look
at appendices unless he/she wants to check something you have done.