Linear vs Logistic Regression
In statistical analysis, it is important to identify the relations between variables concerned to the study. Sometimes it may be the sole purpose of the analysis itself. One strong tool employed to establish the existence of relationship and identify the relation is regression analysis.
The simplest form of regression analysis is the linear regression, where the relation between the variables is a linear relationship. In statistical terms, it brings out the relationship between the explanatory variable and the response variable. For example, using regression we can establish the relation between the commodity price and the consumption based on data collected from a random sample. Regression analysis will produce a regression function of the data set, which is a mathematical model that best fits the data available. This can easily be represented by a scatter plot. Graphically regression is equivalent to finding the best fitting curve for the given data set. The function of the curve is the regression function. Using the mathematical model the usage of a commodity can be predicted for a given price.
Therefore, the regression analysis is widely used in predicting and forecasting. It is also used to establish the relationships in experimental data, in the fields of physics, chemistry, and in many natural sciences and engineering disciplines. If the relationship or the regression function is a linear function, then the process is known as a linear regression. In the scatter plot, it can be represented as a straight line. If the function is not a linear combination of the parameters, then the regression is non-linear.
Logistic regression is comparable to multivariate regression, and it creates a model to explain the impact of multiple predictors on a response variable. However, in logistic regression, the end result variable should be categorical (usually divided; i.e., a pair of attainable outcomes, like death or survival, though special techniques enable more categorised information to be modelled). A continuous outcome variable may be transformed into a categorical variable, to be used for logistical regression; however, collapsing continuous variables in this manner is mostly discouraged because it reduces the accuracy.
Unlike in the linear regression, towards the mean, the predictor variables in logistical regression don’t have to be compelled to be linearly connected, commonly distributed, or to have equal variance inside every cluster. As a result, the relation between the predictor and outcome variables isn’t likely to be a linear function.
What is the difference between Logistic and Linear regression?
• In linear regression, a linear relation between the explanatory variable and the response variable is assumed and parameters satisfying the model are found by analysis, to give the exact relationship.
• Linear regression is carried out for quantitative variables, and the resulting function is a quantitative.
• In the logistic regression, data used can be either categorical or quantitative, but the result is always categorical.