The Logistic Regression Model

987 Words4 Pages
Logistic regression model has been widely used for researching issues in many industries, especially in predicting probability of bankruptcy and default risks of corporations and clients (in banking field). Among those researches, the work Ohlson (1980) could be considered to be outstanding. In his paper, Ohlson used logistic regression model to quantitatively measure and predict the probability of bankruptcy. His trend of work was later studied and upgraded by Zavgren (1985) and Zmijewski (1984). Although the multivariate discriminant analysis (MDA) was the dominating method in studying such fields, regression model arose with many advantages compared to the MDA. The logistic regression model is more appropriate for circumstances where the…show more content…
There are several advantage of such model compared with other traditional one. Logistic regression does not make several key assumptions of linear regression and general linear models based on ordinary least squares algorithms (i.e. linearity, normality, homoscedasticity, and measurement level). Firstly, logistic regression can handle all sorts of relationships between dependent (outcome or response) variable, and independent (predictor or explanatory) variables, or sometimes called covariates, because it uses a non-linear log transformation to the predicted odds ratio. Secondly, the independent variables do not need to be multivariate normal, and even the error terms as well. Logistic regression model does not require the variances to be heteroscedastic for each single predictor as other types of models. And last, the independent variables in Logistic regression model are believed to be able to vary from ordinal, nominal, measures while the model itself could handle the measurement…show more content…
First of all, dependent variables are expected to be binary. There is a drawback for this point, since reducing an independent variable to dichotomous level would obviously lose lots of information. Secondly, as the logistic regression model conducts the response variable in form of probability of the event occurring, it is essential to code the variable accordingly. Thirdly, the model should be fitted correctly with meaningful independent variables, while also all meaningful variables should be included. A good approach to ensure such condition is using a stepwise method to estimate logistic regression. For next assumption, logistic regression requires each observation to be independent, or in other words, the model should have little or no multicollinearity. However, there is the option to include interaction effects of categorical variables in the analysis and the model: if multicollinearity is present centering the variables might resolve the issue, i.e. deducting the mean of each variable. Fifthly, logistic regression assumes there is a linearity of independent variables and log odds. According to the Journal of Statistics Solution Internet-based Organization, whilst it does not require the dependent and independent variables to be related linearly, the independent variables are required to be
Open Document