In this video we will be discussing about concordant pairs discordant peasant type pairs so the concept of concurrent page discordant present itis can be explained with the help of the following table cases where a probability of an event is greater than probability of a non event here even wins yes and on evenings no so probability of yes or probability of an event is greater than probability of non event that is no are called concurrent past cases where probability of event is less than gravity of a non event that is probability of yes is less than probability of no are called discordant pairs and cases where probability of an event the probability of E is equal to probability of no or call type. So let us understand the concept using the table this is the observed outcome This is greater probability. So see here probability of yeses, point seven five and probability of noes point three fifths a week.

This pair so observation one and two together forms concordant pair because probability vs is greater than probability of no then if we take the observation one and observation three we find the probability of no one here is point eight five and probability of yeses point seven five. So just because probability of no is greater than probability of yes it is point eight five is greater than point 75.75. So observation one and three in this together forms a discordant pairs and discordant pairs are bad for the model because it will lead to miss classification and in case of observation, one and observation four, that is observation one and observation for the probability of yeses, 177 probability of nose also conservative, that is probability of yes is equal to probability of no because this position one and four, we call it as Taipei's. Now let's move to the steps which we need to follow to calculate the concordance or discordance.

The first step is calculate the predictive probability logistic regression model divided it into two data sets and data set contains observations having actual value of dependent variable with value One that is event and corresponding predicted priority values and the other data set contains observations having actual value of dependent variables zero which is non event against the predicted probability scopes now let's compare each bracket value in first data set with each rated value in second data set so total number of pairs will be x into y so it's a Cartesian product. So number of observations in first data set is actual values of one independent variable and y is number of observations in second data set actual values of zero independent variable one means even zero means not even now, we are performing partition product that is cross journal events and non events for example, if you have hundred events and thousand non events it could create hundred and 2000 pairs for comparison so that is a Cartesian product is x into y next step a pair is called concordant.

If one that is observation with the desired outcome that is event has a higher probability than zero observation without outcome that is non event a pair is called discordant if zero that is observation without the desired outcome that is known event has a higher credit probability then One observation with outcome that is event and a pair is called tied if one observation with a desired outcome that is event has seen greater probability than zero priority that is observation without the outcome that is known even to the final percent values are calculated using the formula below that is we have to calculate percentage concordant percentage discordance and percentage tight percentage concordant is number of concurrent peers by total number of pairs in 200. percentage discordance is number of discordant pairs the total number of pairs in 200 percentage tied is number of type pairs by total number of pairs in 200.

An area under the curve that is MC is equal to percentage of concurrent plus point five into percentage of type two we will be learning this much in this video for now, let's stop the video here. Goodbye. Thank you. see you for the next video.