Now in this video we will be predicting the probability for y equal to even for every customer or for every individual where each observation denotes one individual who has applied for loan and this criteria probability will be created in a separate data set. So let's do that. Let's predict the probability for equal treatment. So we will be using the procedure proc logistic data. I hope you'll remember that before we run this procedure proc registry data we have to execute our live name statement to bring our data set in SAS environment. So I have shown you on how to execute a lipstick in our last video.

So the live name statement is already executed over here to bring the data set in the SAS environment. So my library name is minus one doc logistic underscore rake underscore German underscore back the S is a keyword to create the model for y equal to one that is equal to even now model responses may dependent variable i will mentioned the significant independent variables or I will specify the significant independent variables that are focus on significant significant independent variables. So let's specify them we have to specify the weight is given in the data set So these are the significant independent variables that is check account, duration, history save account, new car education guaranteed Neil single other installed, installed rate, amount used car, foreign and rent. These are the significant independent variables Now I'm using the statement output Output out out is the key word that he will be using to create the duplicate data set.

My duplicate data set name is resolved which will be created in Zed work because I have not specified in library name p equal to predict it which is used to predict the probability for Viper to event then run so now let's run this code. Before I run the code Let me explain you all the code. So we have used the procedure proc logistic details equal to library name dot data set name. d s is a key word to build a model for rivals to one that is very close to even modern response. As our dependent variable equal to these are the list of significant independent variables which we have obtained in our last result viewers, that is during stepwise selection these are the significant independent variables. That is check account duration history save account new car education character a single other install install rate amount used car foreign rent, we are creating a duplicate data set that is resolved where the predicted probabilities for y equal to one will be displayed.

So the predicted probabilities that is we're predicting the probability for y equal to one that parity priority will be displayed in the data set called resolve which is created in network as I did not specify any library name it will create interwork and then run so now let's run this code. To see here we got the table for percentage concordance percentage discordant percentage tight higher is better percentage amount of concordance better is our model that is less is the risk classification This is odds ratio estimates table this is my analysis of maximum likelihood estimates table that is maximum likelihood estimation is used to estimate the parameters for logistic regression. Okay we are building a model for Y equals to one and as we know that the number of levels for a response variable will be two because response variable is binary in nature. taking two values either zero or one zero denotes non event and one denotes even now let's see the data set that we have created that is resolved.

It is created inside work So, let's open the library work. It's open that he does it result. So, the result is a copy of the original data set that is logistic German bank. So then add it as it got popped in reserve. In addition to that, we have also got another variable that is the estimated growth This is the bank depravity. For Bibles to one that is a greater priority for each customer provide most to one arrivals treatment is distinct.

So, for each customer customer what is the estimated probability that the loan can be given to the customer or the customer will not be alone default is displayed over here this is estimated probability This is listed for all the thousand observations, the thousand observations or the thousand customers who had applied for the loan now, we will be learning this much in this video. Before we move to the next video let me recap the concepts that we have done in the last videos we are using an interest magnetics dataset where we had built a model for Y equals to one and y equals to one denotes that the customer will not be alone default default that that is loan can be given to that customer we had done supply selection for our model. To select the significant independent variables there are total independent variables out of which 14 independent variables are significant ones.

The significant variables are chosen using residual chi square test variation. Notice the model does not require any more variables and each one is a model requires more variables. This is the null hypothesis and alternative hypothesis for a suitable chi square test. We had also we also did Harshman MC goodness of fit test to check the goodness of fit for the model where we concluded that the model is a good fit because our P value is greater than our level of significance that is 0.05. Because the tables are all issue estimates we got the table for analysis of maximum likelihood estimates we also got the table of percentage of components discordance or type s were higher is the percentage of concordance better is our model that is less than his classification. We also generated a classification table for every level of probability from zero to one with a gap of 0.01.

And we got the different measures of classification table that is, correctly classified events correctly classified non events in Korea classified events if the president non events total amount of percentage correctly classified then sensitivity specificity false positive false negative for every level of probability from zero to one, we had created a data set called result where we had predicted the probability for Y equals to one for each customer. That is weird. predict the gravity for each customer, we are predicting the probability for Y equals to one for each customer means that what is the probability that the each customer can be given the loan or the customer will not be a loan default what is the probability that we had predicted and it was, it was displayed in the data set result to the plaintiff gravity for y equal to one is displayed in the data set result which is created in network now in our upcoming videos, we will be setting up a cutoff probability level to convert this estimated priority to a variable called status which is going to take binary value that is zero or one by setting up the cutoff gravity level we will be converting them estimated probability variable to a status variable which will take values zero or one and then we will be doing the confusion matrix between the observed response variable and the predicted response variable which will be a status variable and then we will be calculating the different measures of the confusion matrix manually For now, let's end this video over here.

Thank you. Good. Well, we'll see y'all for the next video.