This week we review statistics from the basics to multivariate statistics. Statistics help researchers to summarize, explore and predict the relations between variables. Basic summarizing tells researchers the general information the data expressed without complex in-depth statistical analysis. Summarizing data can be achieved through measures of central tendency, kurtosis, variability, skewness, relative position, and grouping analysis. However, if we want to measure the association between several variables, we should use more complex statistics, such as regression. 

Regression modelling examines the relations between dependent and independent variables. Linear regression is linear in coefficient; it is the simple regression. Ordinary least square (OLS) is a type of linear least squares for estimating the unknown parameter. It assumes the analysis is fitting a model of the linear relationship between one or more explanatory variables and a continuous or at least interval outcome variable that minimizes the sum of square errors. 

However, the OLS does not analyze the spatial correlation. The geography’s first law argues everything is related and near things are more connected to other things. So to consider the spatial autocorrelation, another regression modelling such as geographically weighted regression is needed by using the distance decay functions. 

Besides, generalized linear models (GLM) is a flexible linear regression of OLS that allows for response variable that has error distribution models other than a normal distribution. It is made up of three components: random, systematic and link function.  

Logistic regression provides an examination for a dichotomous response variable and numeric/categorical explanatory variables by calculating the binary probabilities. The multiple logistic regression extends to more than one predictor variables. It can solve limited dummy dependent variable in some research topics.