WEBVTT 00:01.200 --> 00:04.170 In this session, we will discuss about statistics. 00:05.160 --> 00:09.030 Machine learning is deeply based on statistics. 00:12.500 --> 00:23.390 Statistics is the science and the art of learning from statistics is the collection, analysis and interpretation 00:23.390 --> 00:24.140 of data. 00:24.380 --> 00:29.780 It is the effective communication and presentation of results, relying on the data. 00:30.860 --> 00:39.920 So as part of statistics, we will try to understand how statistics is used to understand the data and 00:39.920 --> 00:42.860 how it could be used to support machine learning. 00:45.650 --> 00:46.100 Here. 00:46.430 --> 00:49.730 You can see that we have a complete population. 00:50.540 --> 00:54.530 Population is the entire dataset which we have. 00:55.280 --> 00:58.760 Let us say we are talking about the entire country. 00:59.860 --> 01:05.770 So the population of the country would be the complete consensus of the country. 01:06.040 --> 01:10.450 All the people present in the country would be a part of the population. 01:11.260 --> 01:18.760 But when we are conducting a particular study, we will not be conducting the survey on the entire population, 01:18.970 --> 01:24.310 but rather be taking a small sample from the population. 01:25.420 --> 01:34.000 But while we are taking samples from the population, the sample would have to be a good representation 01:34.270 --> 01:35.620 of the population. 01:36.130 --> 01:45.220 So the sample has to be selected randomly from the population so that it can represent the viewpoint 01:45.220 --> 01:46.780 of the population correctly. 01:48.790 --> 01:55.330 We use the samples and analyze the samples to derive inferences. 01:55.540 --> 01:55.860 The. 01:57.120 --> 01:58.680 Regarding the population. 01:59.880 --> 02:00.900 This is what? 02:01.920 --> 02:06.150 Population and sample has a relationship between. 02:08.900 --> 02:09.170 Now. 02:09.170 --> 02:11.600 What is population and what is something? 02:12.440 --> 02:16.670 A population includes all the elements from a set of data. 02:17.360 --> 02:23.930 While a sample consists of one or more observations drawn from the population. 02:26.460 --> 02:33.360 A measurable characteristic of a population such as the mean standard deviation is called a barometer. 02:34.050 --> 02:40.560 So all the measurable characteristics of a population are called the barometer, while the measurable 02:40.560 --> 02:44.040 characteristic of the sample are called statistics. 02:45.490 --> 02:49.180 The reports are a true representation of the opinion. 02:49.420 --> 02:56.740 So if they are trying to find out the mean of the population, then the mean of the population is a 02:56.740 --> 02:58.900 true representation of the opinion. 02:59.560 --> 03:08.320 While if we find out the meaning of the sample, then it will have a margin of error and a confidence 03:08.320 --> 03:08.920 interval. 03:09.280 --> 03:13.060 The value will not be exactly same as the population. 03:13.330 --> 03:22.360 There is a possibility of error and there is a confidence interval or interval in which the value of 03:22.360 --> 03:25.420 the mean of a sample can range. 03:26.350 --> 03:32.590 So the mean of the sample will be present in the range between the mean of the population. 03:33.430 --> 03:39.010 So we cannot say that the mean of the sample will be exactly same as that of the population. 03:39.250 --> 03:47.260 It will be a little, but this is why we are applying different methods on the samples so that we can 03:47.260 --> 03:50.560 find out the values of the population accurately. 03:50.830 --> 03:53.350 And this entire thing is called statistics. 03:53.530 --> 03:58.180 The entire study of this characteristic is called statistics. 04:00.490 --> 04:06.730 So why we are taking the samples, we have to apply several sampling techniques. 04:07.870 --> 04:11.320 So there are two types of sampling techniques which are available. 04:11.890 --> 04:17.770 One is the probability sampling, and another one is the non probability sampling. 04:18.970 --> 04:27.250 Probability sampling involves the random selection, allowing you to make statistical inferences about 04:27.250 --> 04:28.210 the whole group. 04:28.930 --> 04:38.170 So in case of probability sampling, the sampling is done by random selection of elements from the population. 04:40.100 --> 04:40.580 Why? 04:40.790 --> 04:50.420 In case of non probability sampling the non random selection based on convenience or any of the criteria 04:50.540 --> 04:54.200 allowing you to easily collect the initial data is used. 04:55.190 --> 05:03.860 So non probability sampling allows a convenient selection of values by probability. 05:03.860 --> 05:08.540 Sampling is used for random selection. 05:08.720 --> 05:15.260 So whenever we are making a random selection it is called probability sampling and whenever the selection 05:15.260 --> 05:20.150 is based on convenience, it is called non probability sampling. 05:22.660 --> 05:26.410 Now there are different types of probability sampling. 05:27.520 --> 05:35.770 One is simple random sampling, where we take random values from the entire set. 05:36.010 --> 05:39.610 We pick out random elements from the entire set. 05:40.180 --> 05:42.640 These are called simple random samples. 05:43.690 --> 05:52.810 Systematic sampling is where we systematically pick a particular sequence of values from the entire 05:52.810 --> 05:53.260 data. 05:53.440 --> 06:00.130 So here, if you see we have values from 1 to 9, so we are picking out all the multiples of three. 06:01.270 --> 06:05.200 To get the samples from this dataset. 06:06.810 --> 06:15.330 The stratified sampling has the values without a clear cluster has formed. 06:15.660 --> 06:23.490 Different clusters have been formed from this data, so cluster of old and young men have been baking 06:23.490 --> 06:24.210 separately. 06:24.600 --> 06:31.740 Another cluster of women have been taken separately and one cluster of children has been taken. 06:32.040 --> 06:40.830 And then from these clusters people are now trying to meet, this is what started this fight, nine 06:40.830 --> 06:48.330 of them sampling this and cluster sampling is we create clusters out of the entire data. 06:48.330 --> 06:56.220 We create different clusters from the data and we just pick out one cluster from this entire data. 06:56.220 --> 07:04.680 We pick out random clusters from this entire clusters of data, and we use them for samples.