Sampling and data: The sampling strategy for the present study is a combination of both purposive and random sampling. The two upazila, namely Jhikargacha and Sharsha under the Jessore district of Bangladesh, were chosen purposively based on the highest flower production area. Four villages were selected from each upazila; whereas the farmers at the village level were randomly selected. A complete list of tuberose growers were collected through the help of DAE personnel from the study areas. Finally, 200 tuberose farmers and 25 farmers from each village were randomly selected and data were collected during the month of February-March, 2017. A participatory methodology was followed to get information about technology adoption followed in the research locations. There are two dimensions of data available for analysis: (i) farmers who have received HYV tuberose, that is, technology receiver and (ii) farmers who did not receive any technology, that is, technology nonreceiver. Analytical technique Data collected were analyzed using descriptive statistics and multiple regression analysis. Descriptive statistics were used to analyze socio-economic characteristics of the farmers and constraints associated with tuberose cultivation. Gross margin analysis was used in analyzing cost and returns in tuberose cultivation per hectare. The tabular method of analysis involved different descriptive statistics and land use cost was calculated based on per year lease value of land. The profitability of tuberose cultivation was estimated using gross margin, net return, and benefit cost analysis. Linear regression analysis Multiple regression analysis was used to determine input output relationship of tuberose farming. Four functional forms, namely, the linear, semi-log, double-log and exponential were tried out while using the ordinary least squares estimates in assessing the regression model. The one that gave the best fit in terms of the magnitude of R2 , Adjusted R2 and the significance of the overall regression as judged by the F-ratio and the significance of the individual coefficients was chosen and reported. The multiple regression model was implicitly stated as: Y = f (X1, X2, X3, X4, X5, X6, X7,.......................................U) (i)
where Y=output of tuberose (stick), X1=Human labour cost ($), X2=Tuberose seedling cost ($), X3=Organic fertilizer cost ($), X4=Inorganic fertilizer cost ($), X5=Irrigation cost ($), X6=Pesticides cost ($), X7=Vitamin cost ($), and U=Error term. Gross margin is the difference between the total revenue (TR) and the total variable cost (TVC). It is a useful planning tool in situation where fixed capital is a negligible portion of farming enterprise as in the case of small-scale subsistence agriculture (Olukosi and Erhaor, 1988; Omotesha et al., 2010; Abdullahi, 2012). Probit model for estimating the determinants of adopting tuberose farming To analyze the adoption of tuberose cultivation, the probit model was utilize. The dependent variables in the adoption model are 0, 1 dummy variables; indicating one if a farmer adopt HYV (double varity tuberose) for flower cultivation and zero if otherwise. According to Gujarati (2004), there are three approaches for estimating the qualitative response of dummy dependent variables: (1) linear probability model (LPM); (2) logit model; and (3) probit model. The linear probability model (LPM) is a typical regression model, but the dependent variable is a dummy variable. The conditional anticipation of the dependent variable, given independent variables, is interpreted as the conditional probability. However, Wooldridge (2009) and Gujarati (2004) dispute that the linear probability model has some negative aspects, including nonnormality of the error term, the probabilities can be less than zero or greater than one, and the partial effect of any independent variable (appearing in the level form) is constant. These confines of the linear probability model can be triumphed over by the logit or the probit model.