06-02-2012 06:08 AM
06-02-2012 06:10 AM
I try to say like this:
....A logistic regression analysis was conducted in JMP ver 9.2 to predict the opportunity of the developer in the high-rise building projects using type of the project, previous land use rights, land rate, rent rate, floor, total investment value, GFA, and campus area as predictors.
A test of the full model against a constant only model was statistically significant, indicating that the predictors as a set reliably distinguished between acceptors and decliners of the offer (chi square = 42.143, p < .0001 with df = 10). RSquare (U) of .26 indicated a moderately relationship between prediction and grouping. Prediction success overall was 90% (92.9% for decline and 87.5% for accept. The Effect Likelihood Ratio Test table demonstrated that only Type, rent price and lease period made a significant contribution to prediction (p = 0.0120, 0.0087 and 0.0116). Previous Ownership, Land_rate, campus, investment_value, GFA, floor were not a significant predictor. EXP(B) value indicates that when type of project is not only office (mix use) the odds ratio is times as large and therefore foreigner developer are also most twice times likely to take the project.
In general, the local developers are different to foreigner developer in type of project, lease period and rent price. The others indicators are not different clearly. In other explanations by odd rations, the probability of the lease period from FDI is shorter but they get the 50% higher rent rate. The other factors (investment value, floor, GFA, land rate..v..v…) is not significant distinctive. Summary, our model will content 3 variables: lease_period, rent_rate and type(0=not only office). Therefore, the our suggested final model is:
The p-value of Intercept is 0.7187, not significant so we do not need it in our model. Moreover, we did not find any outlier in the independent variable. Especially, when all of them transferred to natural logarithm (ln), variables are normal distribution. The Receiver Operating Characteristic (ROC) plots the true foreigner probability vs. the local probability. As the sensitivity increases the false positive rate increases as expected. A perfect classification rule based on upon a logistic model should have area beneath the ROC curve of .90 or higher (Rang 2009). Here we do not quite meet that standard but AUC=0.82638 is still good distinguish ability.
From the above figure, the foreign investor strongly forcus on the big commercial projects such as mix-use, hotel. While the local investors are very flexible on the quantity of office building. The aim of the 100FI aims to the foreigner costumers want to open their branch office in this new developing market.
Is it correct or wrong?