In many real fields such as biostatistics, survival analysis, epidemiology, sociology, logistic regression model is used when the response variable is of dichotomous or polytomous category. The logistic regression model may consist of several discrete and continuous covariates. When logistic regression model consists of continuous covariate the model should posses the assumption that there is a linear relationship between the logit and the continuous covariate. When the assumption is violated the analysis of cause-effect relationship via logistic regression will not be appropriate and may mislead the results. So, for assessing the actual effect of the continuous covariates some transformations of the respective covariates are used to establish linear relationship between the logit and covariates. This research deals with the aspect of linearity assumption between logit and continuous covariate while analyzing Bangladesh Demographic and Health Survey (BDHS) child mortality data with the help of logistic regression. The aim was to see how the actual effect of a covariate is distorted if the linearity assumption between the logit and covariate is violated. Also the thesis showed how the covariates are transformed to reach the linearity assumption. Various transformations such as power, logarithmic have been used to make that relationship linear. The odds ratios, for the respective transformed covariates, have been calculated and accordingly the effects of the new transformed covariates have been examined through the new odds ratio formula.