Course teached as: B032747 - STATISTICS FOR EXPERIMENTS AND FORECASTS IN THE FIELD OF TECHNOLOGY Second Cycle Degree in MANAGEMENT ENGINEERING
Teaching Language
STATISTICS FOR EXPERIMENTS AND FORECASTS IN THE FIELD OF TECHNOLOGY (R.BERNI-6 CFU)
ENGLISH
------
FORECASTING METHODS (A.Magrini-3CFU)
ENGLISH
Course Content
The course is composed of two parts:
STATISTICS FOR EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R.BERNI-6 CFU) presents the design of experiments (DoE) and process optimization for the field of technology;
FORECASTING METHODS (A. Magrini -3CFU) presents fundamental methods for time series forecasting.
*** For the students of the Master Degree in Statistics and Data Science”, see the section “Further information”
STATISTICS FOR EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R.BERNI-6 CFU)
1-Montgomery DC, 1991, Design and analysis of experiment, Wiley- (chapter 4,5, 6,8; 10; 11 -fino a 4.2 incluso-).
2-Khuri I e Cornell JA, 1987, Response surfaces: design and analyses, Ed. Marcel Dekker- (chapter n. 1, 4, 5).
3-Berni R.,2014 working paper elettronico n.10;
http://local.disia.unifi.it/wp_disia/2014/wp_disia_2014_10.pdf
Please note
that all the cited Textbooks are available at the Universitary Library – Polo Scienze Sociali di Novoli.
Other books for alternative readings:
Introduzione alla Statistica Applicata con esempi in R. Federico M. Stefanini, Ed. Pearson.
Probabilità e statistica per l'ingegneria e le scienze; Sheldon M. Ross; Ed.Apogeo.
------
FORECASTING METHODS (A. Magrini - 3CFU)
• Wayne A. Woodward, Henry L. Gray, Alan C. Elliot (2017). Applied Time Series Analysis with R, 2nd edition, CRC Press.
**consultabile**:
Søren Bisgaard, Murat Kulahci (2011).
Time Series Analysis and Forecasting by Example, John Wiley & Sons.
• Additional resources supplied by the teacher about: R programming; Analysis and treatment of outliers.
Learning Objectives
STATISTICS FOR EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R.BERNI-6 CFU)
To develop knowledge and abilities in order to analyse data through statistical models, by also considering peculiar issues of the basic statistics , and also in the technological field. To perform an efficient experimental planning, to optimize a product or a production process with respect to specific characteristics (target) of quality and/or reliability (cc1, ca1). The latter also considering the phenomenon under investigation, the real context (external source of variabilities, noises). The possible decisional and technical implications (cc7) are also taken into account.
To understand capabilities and limitations of the methods (cc8) in order to suitably join theory and practice (ca5) through the application of the theory to real data, and correctly performing the potentialities of the methods with respect to environment and the process to be studied (CT5, CT10).
To develop own abilities, starting from the design planning step up to the final optimization step, by also considering the robust process optimization (CT7).
To develop analytical and critical abilities, so as to try refinements or explore different methodologies depending on the characteristics of the data (self-learning) (CT5, CT7)
------
FORECASTING METHODS (A. Magrini -3CFU)
To develop knowledges and abilities to make predictions, under uncertain conditions, using time series data (cc1, ca1). The phenomenon under investigation, the applied context and the possible decisional implications (technical – cc7 and not – cc9) are also taken into account.
To understand capabilities and limitations of the methods (cc8) so as to link suitably theory and practice (ca5; ct5, ct10).
To develop judgment and communication abilities, also using English (see the section “Modalità di verifica dell’apprendimento”; ct1, ct3, ct6).
To develop analytical and critical abilities, so as to try refinements or explore different methodologies depending on the characteristics of the data (self-learning).
Prerequisites
Calculus, linear algebra, probability calculus, statistical inference (estimation, confidence intervals, significance tests), linear regression model, programming language abilities.
Teaching Methods
STATISTICS FOR EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R.BERNI-6 CFU)
Traditional lesson through slides and blackboard; in some specific cases, a pen tablet could be used (slides delivered after the lesson) and practice in the computer lab. Case studies. Individual task (or in group) on real data.
------
FORECASTING METHODS (A. Magrini - 3CFU))
Traditional lesson (slides delivered after the lesson) and practice in the computer lab. Alternatively, each student can use his own laptop.
****** IMPORTANT NOTE ONLY FOR THE STUDENTS OF THE MASTER DEGREE IN STATISTICS AND DATA SCIENCE: for the 3 CFU part to be taken with the teacher Rossella Berni, contact her for precise indications on the beginning and the precise dates of the lessons to be attended*****
Further information
Use of Moodle platform
*****IMPORTANT NOTE ONLY FOR THE AY 23-24 AND ONLY FOR THE STUDENTS OF THE MASTER DEGREE IN STATISTICS AND DATA SCIENCE:
The course B032747-STATISTICS FOR EXPERIMENTS AND FORECASTS IN THE FIELD OF TECHNOLOGY borrows the course
B031816 - STATISTICAL METHODS FOR FORECASTING AND QUANTITATIVE MARKETING (Master Degree in Statistics and Data Science). To this end:
- STATISTICS OF EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R. Berni-3CFU): the students will be asked to do further ad hoc topics related to choice experiments, to be defined with the teacher;
- FORECASTING METHODS (A. Magrini-3CFU): the students will be asked to do further ad hoc topics related to statistical modelling for estimating dynamic causal effects and forecasting multivariate time series, to be defined with the teacher.
****
Type of Assessment
STATISTICS FOR EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R.BERNI-6 CFU)
Oral examination. Questions will be related to the main arguments of the course, as described in the diploma supplement. Particular attention is payed to the critical and constructive student's abilities, e.g.: the student must establish to be able to deeply discuss the topics of the course (CT5).
More precisely, in addition to the theory, the student should be able to apply the studied methods, by also showing through examples (also using the case-studies illustrated during the teaching course) the empirical aims of the course (CT10). Particularly appreciated the ability to discuss and compare the alternative methods (example: to be able to compare the split-plot planning with respect to a CCD in an empirical situation, also introducing simple cases) (CT7, CT10).
------
FORECASTING METHODS (A. Magrini-3CFU)
Exam composed of two parts:
1) Empirical analysis of real time series with the preparation of a written report (ct1, ct5, ct6, ct10) (30% of the final grade);
2) Oral examination with a short discussion of the report (ct3) and theoretical questions (70% of the final grade).
-------
The final vote (9 CFU) is computed as the weighted average of the two parts
**The exam FORECASTING METHODS (A. Magrini-3CFU) must be carried out before the exam STATISTICS FOR EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R. Berni-6CFU)**
--------------------------------------------------------
For the students of the master Degree in Statistics and Data Science, the final vote (6 CFU) is computed as the average of the two parts
Course program
STATISTICS FOR EXPERIMENTS IN THE FIELD OF TECHNOLOGY (R.BERNI-6 CFU)
The course (6 CFU) is slightly modified in the initial part w.r.t. the previous AA years. Now it is based on three parts, strictly related to each others; these connections will be illustrated along the course.
PART #1
the concept of: variable and factor; qualitative and quantitative variable, discrete and continuous; modality and level. Linear regression model simple and multiple; parameter estimation; OLS; covariance; correlation coefficient; diagnostic measures; R-square and adjusted. Examples, residual analysis. Test of significance of regression. Test on individual regression coefficient and groups of coefficients. Confidence interval and prediction interval. Test Lack-of-Fit (also in Part #3). In the laboratory sessions (in R): examples with qualitative and quantitative independent variables (dummies). Examples and relation between the linear regression model and Analysis of Variance- ANOVA.
PART #2
ANOVA (one-way and two-way); difference between observational data and experimental data; design of experiment (DoE): basic elements (replication, source if variability, randomization). Randomized Complete Block Design.- RBC, Latin Square design; experimental and sub-experimental factor: noise and block; interaction among factors; 1st order interacion; full factorial design and fractional factorial design at 2 levels (Resolution criterion, counfounding, aliasing, alias pattern, word defining relation, the building of the fractional factorial design). During the laboratory sessions: ANOVA, the model strategy in ANOVA; ANOVA with a fractional factorial design, p-value classification: p-value null or negligible.
PART #3
Response Surface Methodolgy-RSM and Split-plot design.
Starting from the local series Taylor approximation around x0, we illustrate the basic principles of RSM (one response variable) considering the sequential feature of RSM in the I order, and also between I and II order.
I order RSM (I order designs- see part#2; 1st order polynomial model; steepest ascent/descent);
Test LOF; curvature test; Pure error and replications in X0.
RSM II order (Central composite design-CCD and Full factorial design at 3 levels).
Split-plot design, basic hystorical design (ANOVA) and split-plot in RSM; random and fixed effects and Mixed Response Surface models.
Design properties in RSM. Optimization methods: canonical analysis and ridge analysis.
During the session labs (in SAS): 2nd order polynomial model applications and model strategy in RSM; Optimization and setting of factor levels for robust process optimization (dual response approach and dual response approach simplified in a multireponse situation).
At the end three real case studies will be illustrated to give a global comprehension of the course.
The Laboratory sessions are included in the examination programme.
------
FORECASTING METHODS (A. Magrini-3CFU)
• Introduction to the R environment
• Introduction to time series analysis and forecasting
• Time series and stochastic processes
• Some examples of stochastic processes: White Noise, Random Walk, AR(1), MA(1)
• Simulations
• Weakly stationary and ergodic processes
• Autocorrelation function (Acf) and portmanteau tests
• AR (Autoregressive), MA (Moving Average) and ARMA processes
• Non stationary time series and integrated ARMA processes (ARIMA)
• ARIMA processes with seasonal components
• Estimation, selection and diagnostics of an ARIMA model
• Forecasting in general and with ARIMA models
• Forecasting error measures and forecast comparisons
• Transformations of variables
• Analysis and treatment of outliers
• Use of external regressors
Sustainable Development Goals 2030
Some case-studies aim at improving the quality of the environment