WEBVTT 00:01.130 --> 00:06.560 In this video we learn how to create joint plots under aggression plots and this these plots we can 00:06.560 --> 00:12.950 graphically analyze whether that's a positive or negative functional relationship between features. 00:12.950 --> 00:15.040 So for example now a titanic Dead set. 00:15.060 --> 00:21.580 There are the features age and fair and we could try to answer the question whether for example overlap 00:21.710 --> 00:25.630 hasn't just pay more than a young gap has themselves for that tickets. 00:25.640 --> 00:31.990 And in this video we will also further graphically analyze whether the age or the fare of the passenger 00:32.000 --> 00:36.280 had influence on the chance to survive the disaster. 00:36.290 --> 00:40.210 So first of all its import pandas see born and met plot slip. 00:40.490 --> 00:45.680 And again we are working with our Titanic data set and explain our Titanic data set. 00:45.680 --> 00:52.760 We have two numerical columns to continuous numerical columns H and fair and in a that we can create 00:52.790 --> 00:59.960 a joint plot which is actually a setup plot where we have on the exact CIS the age and on the y axis 01:00.080 --> 01:03.560 be fair and kind of the plot. 01:03.560 --> 01:07.180 We can pass yes get the plot so you want get a scatter plot 01:11.430 --> 01:17.670 and we can see on the x axis that the AIDS and on the y axis the fair and each single the dots here 01:17.670 --> 01:19.880 are each single point to say a passenger. 01:20.190 --> 01:25.490 And here at the top end on the right hand side we can also see the frequency distribution. 01:25.860 --> 01:32.370 So this is the frequency distribution or histogram for the fair and this is the the histogram for the 01:32.370 --> 01:33.030 H. 01:33.510 --> 01:38.820 So we can see that the majority of ages are here around between 20 and 40. 01:39.300 --> 01:43.410 And for the fare here around between zero and 50. 01:43.680 --> 01:49.800 And also here we have a future warning which is caused by an intersection between the side pi packets 01:50.190 --> 01:51.720 and the Seabourn package. 01:51.720 --> 01:55.750 So we can ignore this here and actually most of the apparently to hide. 01:55.770 --> 02:04.590 We can determine here the size and the height of the graph and we can also change the kind of the plot. 02:04.590 --> 02:07.770 So instead of having a scatter plot we can have different plots slow. 02:07.770 --> 02:09.120 Let's check this here. 02:12.060 --> 02:19.730 A.B. has the option to pass get a regression or is the delta KDE or hex so let's pass here hex 02:25.160 --> 02:29.820 and hex stands for a hex bin plot and here we can see that our blood are here. 02:29.840 --> 02:36.530 The area is divided into two equal size Texas guns and the darker the color of the aspect affects like 02:36.530 --> 02:39.560 on the more data points we have here. 02:39.590 --> 02:45.230 So here we can see the majority of data points that we have here at the quiet. 02:45.320 --> 02:54.810 Low fat prizes and at ages between 20 and thirty eight or something and we can also pass here K D E 03:00.600 --> 03:07.570 and Katie U stands for kind of density estimates or so he has set off having a histogram BFM kernel 03:07.620 --> 03:15.390 density estimate or kind of a line that approximates the histogram and you know a graph you can see 03:15.390 --> 03:23.790 actually some areas and the darker the areas the more data points we have fear so he can identify where 03:23.850 --> 03:26.920 the majority of data points are located. 03:26.920 --> 03:29.460 All right now let's go back to the scatter plot 03:36.090 --> 03:42.150 so here we can see the scatter plot between H and fair and the next question would be whether there 03:42.150 --> 03:46.340 is a significant relationship between age and fair. 03:46.350 --> 03:52.590 So for example of the old hasn't just paid more than younger passengers and we can answer this question 03:52.590 --> 03:57.630 by having a regressive plot and therefore we can also pass the Rick 04:02.690 --> 04:09.350 and TB can see a regression line through the data points and the regression line seems to be upward 04:09.350 --> 04:17.080 sloping so here we have a positive relationship or a correlation between the H future and the fast feet. 04:17.540 --> 04:23.870 So the older the peasants are the higher the ticket price and you can also see here the shaded area 04:24.110 --> 04:31.880 around the regression line and this marks actually the a 95 percent confidence interval of the regression 04:31.880 --> 04:32.560 line. 04:32.660 --> 04:39.970 So the regression line itself is actually a point estimate and let's assume that V is statistically 04:40.220 --> 04:48.170 similar to the booking for the Titanic 1000 times then in 95 percent of all cases the regression line 04:48.200 --> 04:51.470 between H and fair is in this area here. 04:51.530 --> 04:59.000 So it seems like that the positive slope or positive relationship is statistically significant here 04:59.600 --> 05:05.980 and seaborne actually provides some alternatives to create a regression plot with some y functionality 05:06.680 --> 05:10.220 and for example we can do this with a l m plot. 05:10.340 --> 05:14.870 So again here we pass them our Titanic data frame to the parameter data. 05:15.080 --> 05:23.780 Then we want to f on the exact S H on the y axis fair and with the parameters aspect and height we can 05:23.780 --> 05:28.040 determine externally the size of the plot or the subplots. 05:28.040 --> 05:30.240 So we will see this in a minute. 05:30.290 --> 05:31.040 Let's run here 05:34.030 --> 05:40.480 so essentially we have actually the same regressive plot with the same regression line and the confidence 05:40.500 --> 05:41.240 in time. 05:42.130 --> 05:43.360 So let's have a look here. 05:48.000 --> 05:53.560 So again here we have parameters like CIA which stands for the confidence interval. 05:54.240 --> 05:59.370 So here in this confidence interval here we have a 95 percent confidence. 05:59.370 --> 06:06.400 And under the hood seaborne actually simulates the booking of the Titanic 1000 times. 06:06.480 --> 06:11.580 And by doing so it creates a confidence interval for the regression line. 06:11.580 --> 06:19.180 So this is the same as above but actually here the alleged plot provides some additional functionality. 06:19.590 --> 06:26.760 And also here we have the parameter hue and we can split up our data frame for example by the sex column 06:32.260 --> 06:39.310 so we can see that we have categorized all single points or dots as being either male or female. 06:39.730 --> 06:47.230 And here we have female with blue collar and for the the dots points we have one regression line and 06:47.230 --> 06:49.980 also for the females we have fun regression line. 06:50.110 --> 06:57.970 And it seems like that for males and females the positive relationship is serious statistically significant. 06:58.000 --> 07:04.910 So we have an upward sloping regression line and instead of having here both the regression lines and 07:05.080 --> 07:12.280 on the graph we can also create subplots and we can do this and the set of passing of the sex column 07:12.280 --> 07:21.090 to the parameters you can posit to the pair meet meat columns Carlo and by doing so they're creating 07:21.090 --> 07:23.940 a two subplots and actually two columns. 07:24.360 --> 07:32.190 So here we have the regression plot for the male peasant us and your father female passengers and both 07:32.460 --> 07:41.380 seems to be actually here upward sloping and theoretically we can also create subplots in rows by passing 07:41.390 --> 07:47.760 year sex to the rows pairing meta. 07:47.830 --> 07:49.700 So now we have two rows. 07:49.750 --> 07:54.790 We have the regression father made a peasants us and for the female peasants us. 07:55.120 --> 07:56.310 But let's stick here. 07:56.320 --> 07:57.180 Two columns 08:00.230 --> 08:06.600 and two bits of the pyramids aspect and height we can actually customize the size of them the subplots 08:06.600 --> 08:07.580 here. 08:07.770 --> 08:13.980 So here we might conclude that there is a positive relationship between H and fair not only for the 08:14.000 --> 08:16.960 male peasant does but also for the female peasant us. 08:17.190 --> 08:23.520 And now we want to analyse the factors that actually influence them the chance to survive. 08:23.520 --> 08:27.100 And again we are creating here an adolescent plot. 08:27.180 --> 08:31.470 Then on the x axis our first factor is the age. 08:31.650 --> 08:38.640 And then on the y axis we have our survival column so far this survival column we know that we only 08:38.640 --> 08:43.130 have the values that 0 and 1 and let's running. 08:43.200 --> 08:52.490 So so here on the x axis we can see the eight and on the y axis we have either zeros or ones. 08:52.490 --> 08:54.690 And here we have a regression line. 08:54.690 --> 09:00.860 But here we have a so-called logistic regression because our dependent variable here. 09:00.870 --> 09:07.600 So the survived by our are but it's actually a categorical feature maybe either half the value says 09:07.620 --> 09:15.180 0 and 1 and therefore we have to set he had the logistic Parramatta to true because here we want to 09:15.180 --> 09:17.610 perform a logistic regression 09:28.290 --> 09:29.190 so now we can see here. 09:29.190 --> 09:36.090 I want to just stick regression line and we can see that it's the downward sloping and I see the confidence 09:36.120 --> 09:41.360 intervals the chest that this might be statistically significant. 09:41.370 --> 09:47.610 So the older the peasant sir the lower so the chance to survive and the vice versa. 09:47.610 --> 09:52.330 So the younger the peasants the higher the survival rate the chance to survive. 09:52.410 --> 09:56.550 And this is actually in line with that women tend to run fast. 09:56.550 --> 10:03.360 So children had a better chance to survive than older passengers and also here we can make a categorical 10:03.360 --> 10:10.020 split so we can split our data frame into groups and therefore we can pass the Flex hopper to the. 10:10.150 --> 10:13.290 Para meta a categorical variable. 10:13.500 --> 10:15.180 So for example the six 10:26.080 --> 10:32.780 so here we have the logistic regression between H and survived form a peasant us and female peasant. 10:33.130 --> 10:36.990 And it looks like that here we have a difference. 10:37.120 --> 10:43.770 So if I'm a peasant says it seems like the angle of the and so the higher the chance to survive. 10:44.170 --> 10:49.330 And for female peasants us the relationship is actually the other way around. 10:49.570 --> 10:52.850 So that could mean that form a peasant does. 10:53.080 --> 10:56.510 It was even more important to be a child. 10:56.620 --> 10:57.120 All right. 10:57.130 --> 11:02.770 Next year we can do the same analysis also here split it by the PE class. 11:02.800 --> 11:10.090 So we have still here on the x axis that the H and on Devi axis these survived column and then we split 11:10.120 --> 11:13.000 our date data frame by that peak plus column 11:15.860 --> 11:16.520 and again here. 11:16.520 --> 11:20.130 This is a logistic regression SVM funded by axis. 11:20.240 --> 11:25.270 The survive feature but the categorical data zero of one. 11:25.280 --> 11:29.510 So we have to pass here true to the logistic parameter 11:32.150 --> 11:34.570 so true with capital T of course 11:40.950 --> 11:48.860 and here we can see the logistic regression for the three passenger classes between H and survived. 11:49.050 --> 11:55.470 And it seems to that for all three passenger classes the relationship holds true that the order the 11:55.470 --> 12:00.560 passenger the lower the chance to survive the disaster. 12:00.680 --> 12:01.100 All right. 12:01.100 --> 12:06.530 This was the the video on regression lines and this is pretty helpful to visualize relationships between 12:06.540 --> 12:07.400 their features. 12:07.670 --> 12:14.960 So now we understand better the factors that influence actually the chance to survive for the Titanic 12:14.960 --> 12:16.150 disaster. 12:16.400 --> 12:18.970 And with this we are finished here and out to sea. 12:18.970 --> 12:20.870 Also in the next video by.