WEBVTT 00:00.910 --> 00:08.200 Now, let's get further deeper into hypothesis testing, so while doing hypothesis testing, we might 00:08.860 --> 00:10.750 come at two types of errors. 00:11.990 --> 00:16.430 The two types of errors are five one error, and they do it. 00:17.500 --> 00:24.700 Here you can see a truth table, so here we have witch's brew to about the population. 00:26.140 --> 00:33.570 That this one could be that the hypothesis null hypothesis vagi, which we have figured out is true, 00:33.760 --> 00:37.510 and another would be that the alternate hypothesis could be true. 00:38.970 --> 00:43.800 Now, based on the samples, we can make two types of decisions. 00:44.070 --> 00:51.570 One is to reject the null hypothesis and another would be to accept the null hypothesis. 00:52.500 --> 01:00.150 Now, in case the truth about the population is that the null hypothesis is true, but based on the 01:00.150 --> 01:04.290 data, we somehow reject the null hypothesis. 01:05.140 --> 01:08.190 Then this is called type one errors. 01:10.430 --> 01:18.890 And another type of error is when the truth about the population is that the alternate hypothesis is 01:18.890 --> 01:21.320 true by the. 01:22.310 --> 01:30.590 Accept the null hypothesis based on the samples given, so when we accept the null hypothesis, when 01:30.590 --> 01:37.250 actually all the data but this is was true, that is called by the data and when the null hypothesis 01:37.250 --> 01:40.940 is true, but we somehow magically it is called the type one. 01:40.940 --> 01:41.270 Never. 01:42.310 --> 01:50.980 Some example of the type one error could be one example where a male human is tested positive for being 01:50.980 --> 01:51.520 pregnant. 01:52.150 --> 01:55.150 Now, this scenario is not really possible. 01:55.990 --> 02:02.820 So formally it is defined as the incorrect rejection of a null hypothesis. 02:03.580 --> 02:09.340 So the null hypothesis in this case will be that a male human is not pregnant. 02:10.240 --> 02:18.910 So the hypothesis is that male human is not pregnant and this hypothesis is true. 02:19.210 --> 02:27.490 The hypothesis is completely true that a human cannot be pregnant, but based on the data, reject this 02:27.490 --> 02:31.030 null hypothesis, which is type one error. 02:32.930 --> 02:42.370 The other type of error is when null hypothesis is that a human is pregnant, a male human is pregnant, 02:42.560 --> 02:46.200 and the best support the null hypothesis. 02:46.490 --> 02:54.530 So when we are saying that a male human is pregnant and the test suggests that this is true and it kind 02:54.530 --> 02:57.600 of allows us to accept the null hypothesis. 02:57.800 --> 03:03.230 So this is called tied Twitter because here they put the hypothesis is not correct. 03:03.440 --> 03:06.390 The null hypothesis is not true in this condition. 03:06.740 --> 03:11.580 So it is defined as the acceptance of the false hypothesis. 03:12.470 --> 03:18.500 So these are the two types of errors which we can commit by performing hypothesis testing. 03:25.600 --> 03:33.160 Now, while we were discussing about hypothesis testing, we were continuously talking about the statistical 03:33.160 --> 03:34.090 significance. 03:34.390 --> 03:37.390 So what is the statistical significance? 03:38.200 --> 03:47.200 The p value is how likely it was that our sample was drawn from a hypothetical population where nothing 03:47.200 --> 03:48.040 was going on. 03:49.120 --> 03:58.260 So it tells us, like how likely it is that the population is where nothing was going on, so that the 03:58.550 --> 04:06.280 statistical significance simply means that the all being preserved are unlikely to be present, a situation 04:06.280 --> 04:08.440 where there was no relationship between the. 04:10.960 --> 04:11.410 So. 04:12.880 --> 04:18.620 The differences are big enough to unlikely to have happened simply due to jobs. 04:19.360 --> 04:28.780 So we want to make sure that the two means which we have achieved are actually not because of John's. 04:29.830 --> 04:37.290 So we want to find out if there is a chance that they are just because of what John saw different, 04:37.510 --> 04:44.590 or is it actually a part of two different population, which is the reason why there is so much difference 04:44.590 --> 04:45.170 between them? 04:45.520 --> 04:53.110 So if the values are significantly Partovi, if the values are far obvious or that one value is in the 04:53.110 --> 05:01.240 central area and another value is found in the fields of the gold or in the critical region or below 05:01.240 --> 05:07.150 or above the region, then it is going to be significantly different. 05:08.860 --> 05:14.740 The smaller the P value, the greater your confidence in the statistical result. 05:15.160 --> 05:22.270 Now, the smaller the probability value, the smaller the probability value of the the value being part 05:22.270 --> 05:28.690 of the same distribution, the more is the chance that it is not a part of the same distribution, but 05:28.690 --> 05:30.910 actually a part of a different distribution. 05:32.030 --> 05:37.970 Now, the fight does not change, whereas the P-value are dependent on the actual value of those statistics 05:37.970 --> 05:44.720 and the question so the value is just the line which we have drawn, it is just a constraint which we 05:44.720 --> 05:51.230 have given that we have won to have ninety five percent level or ninety nine percent level or 90 percent 05:51.230 --> 05:51.560 level. 05:51.860 --> 05:58.430 If value will not change by the probability of the value which we are actually trying to find out on 05:58.430 --> 06:03.050 the namiko, that value will keep on changing. 06:03.470 --> 06:10.880 That is something which we are trying to find out if it is significantly far away from this original 06:10.880 --> 06:16.600 value or is it not part of a belongs to the same namiko? 06:21.500 --> 06:28.280 In the next session, we will work on this and we will try to understand what the best is and how we 06:28.280 --> 06:32.660 can conduct this and what are the different types of best.