**Tools and Issues in Data Collection and Analysis (****Third part)**

** **April-June 2014, Palazzo Strozzi

Dr. Federico Russo

The third part covers the basic concepts of inferential statistical analysis and introduces the classical linear regression model. The recommended text for the third part (final set of nine encounters) is:

*Alan Agresti and Barbara Finlay, Statistical Methods for the Social Sciences (3rd or 4th Edition), Pearson.*

The more mathematically inclined students may find useful also the following book:

* Damodar N. Gujarati, Basic Econometrics with Applications, (4 ^{th} o 5^{th} Edition) McGraw−Hill.*

**23 April – 10-12:30 p.m. (SIENA)**

**Introduction to the third part**

Statistical Methods are increasingly employed in Political Science to test hypotheses about social and political phenomena. The growing power offered by computers and simple statistical packages opened new analytical possibilities, but are no substitute for a firm understanding of the basic inferential techniques.

**23 April – 2-4:30 p.m. (SIENA)**

**Probability distributions**

Inferential statistical methods use sample *statistics* to make predictions about the values of some *parameters* of the population of interest. To understand how this is done it is essential to introduce the concept of probability and sampling distributions.

*REQUIRED READING: Chapter 4. Probability Distribution (Agresti & Finaly, 4 ^{th} edition)*

**7 May – 10-12:30 p.m.**

**Estimation
**

Sample data can be used to form two types of estimator of parameters, a *point estimate* and an *interval estimate*. Both can be estimated for quantitative variable (means) and for qualitative variables (proportions).

*REQUIRED READING: Chapter 5. Statistical Inference: Estimation (Agresti & Finaly, 4 ^{th} edition)*

**14May – 3-5:30 p.m.**

**Significance test**

Theories generate hypotheses. A common aim in many studies is to check whether the hypotheses generated by a theory are compatible with the empirically observed data. This can be done with two complementary approaches, the *significance test *and *confidence interval* approach.

*REQUIRED READING: Chapter 6. Significance Tests (Agresti & Finaly, 4 ^{th} edition)*

**21 May – 3-5:30 p.m.**

**Introduction to the two variable regression model**

The two variable regression model studies whether an association exists between two quantitative variables, the strength and the form of that relationship.

*REQUIRED READING: Chapter 9. Linear Regression and Correlation (Agresti & Finaly, 4 ^{th} edition)*

** 26 May – 3-5:30 p.m.**

**When there are several Independent Variables: Multiple Regression**

It is often necessary to go beyond bivariate analysis to study partial relationships between two variable controlling for other variables. The multiple regression model allows for that.

*REQUIRED READING: Chapter 11. Multple Regression and Correlation (Agresti & Finaly, 4 ^{th} edition)*

** 28 May – 3-5:30 p.m.**

**ANOVA models**

Qualitative explanatory variables often play an important role in political theories. For quantitative response variables, ANOVA model is a way to compare the mean responses of several groups defined by the categories of the qualitative explanatory variable.

*REQUIRED READING: Chapter 12. Comparing Groups (Agresti & Finaly, 4 ^{th} edition)*

** 4 June – 3-5:30 p.m.**

**ANCOVA models
**

When there are both quantitative and qualitative explanatory variables regression and ANOVA must be combined.

*REQUIRED READING: Chapter 13. Combining Regression an ANOVA (Agresti & Finaly, 4 ^{th} edition)*

**11 June – 3-5:30 p.m.**

**Issues and tools in model building**

Building a regression model involves various steps that are often neglected, such as checking regression assumptions and take remedial actions when some of them are not entirely satisfied.

*REQUIRED READING: Chapter 14. Model Building with Multiple Regression (Agresti & Finaly, 4 ^{th} edition)*