Global Journal of Management and Business Research, E: Marketing, Volume 22 Issue 3

with the sample size of 99 respondents from customers in CRDB Chamwino Branch. = N 1 + N(e) 2 Whereby n = Sample size, N = Targeted population, e = Level of precision or confidence interval i.e., 10%. The reasons of adopted 10% and not 5% or 1% is due to the coefficient variation of the population within the researcher interest. = 12,670 1 + 12,670(0.1) 2 = 12,670 127.7 = a) Econometric Model Specification The study used binary logistic regression model to approximate the chances of the binary variable with two probable result events such as pass/fail, win/lose, high/low as recommended by (Ozsari and Food, 2016). The customer satisfaction is the discrete random variable and dummy in scenery that could be measured through binary logistic regression or logit model. ( ) ( ℎ) ( ) =1 ( =0) ′ ……………………………………… (1) IV. F indings and D iscussion Therefore, internal reliability of the 18-item scale was assessed. Results indicated a reliability alpha value of 0.965. This indicates that the internal consistency of items is to the extent of 0.965 out of 1 indicating a very high and reliable consistency of the items. Table 1. indicates the findings. Table 1: Reliability Statistics for All Items Variable Number of Items Alpha Value Cronbach`s Alpha Value 18 0.965 Source: Research Findings (2022) a) Binary Logistic Regressions Analysis A binary logistic regression analysis was carried out to estimate the logit model. Since the responses of a dependent variable (Customer Satisfaction) had 5-point Likert scale responses; then a cutoff point of was created where all the mean values of 3.5 and above represented customers who are satisfied and were given a value of one. 1. while the rest represented unsatisfied and were assigned a value of zero (0) to make binary logistic regression possible. b) Binary Logistic Regression Goodness of Fit Test The Hosmer and Lemes how test were used to run the goodness of fit test for the model. Therefore, the Hosmer & Lemeshow test (Table 2) of the goodness of fit proposes the modelisa good fit to the data as p=0.279 whichis greater than 0.1. Table 2: Hosmer and Lameshow Goodness of Fit Test Step Chi-square Df Sig. 1 9.810 8 0.279 Source: Research Findings (2022) c) Omnibus Test of Model Coefficients The omnibus test of model coefficients tests whether the model is statistically significant and can further be interpreted. From the fact that the model has a p-value of 0.000 (Table 4.2.2) which below 0.05 this suggests that the model is statistically significant and can further be used for estimations since the overall model is statistically significant; χ 2(3)=59.465, p <0.05as indicated on table 4. Table 3: Omnibus Test of Model Coefficients Step Chi-square Df Sig. 1 9.810 8 0.279 Source: Research Findings (2022 ) 19 Global Journal of Management and Business Research Volume XXII Issue III Version I Year 2022 ( ) E © 2022 Global Journals Contribution of Mobile Banking Informational Service on Customer Satisfaction in Tanzanian Commercial Banks