1 00:00:01,667 --> 00:00:04,142 Hi. In this video we are going to continue 2 00:00:04,142 --> 00:00:05,759 to talk about inference. 3 00:00:05,759 --> 00:00:09,446 But now we're going to be talking about how you can conduct hypothesis 4 00:00:09,446 --> 00:00:12,427 tests in R. 5 00:00:12,427 --> 00:00:15,660 So the general hypothesis testing procedure 6 00:00:15,761 --> 00:00:19,651 is we always state hypotheses about your parameter. 7 00:00:20,308 --> 00:00:23,288 We collect some data. 8 00:00:24,147 --> 00:00:27,128 We construct a test statistic. 9 00:00:29,249 --> 00:00:33,594 We then apply a decision rule so we can either 10 00:00:34,604 --> 00:00:37,585 do that through a critical value 11 00:00:37,736 --> 00:00:41,222 or with p-values or like a critical region, excuse me. 12 00:00:41,222 --> 00:00:43,849 Or with p-values. 13 00:00:43,849 --> 00:00:45,668 And then we will draw 14 00:00:45,668 --> 00:00:48,648 conclusions in context. 15 00:00:49,254 --> 00:00:52,689 So the first research question we're going to talk about 16 00:00:52,689 --> 00:00:56,983 today is we're going to continue using the idea of iris flowers. 17 00:00:57,438 --> 00:00:59,358 And see like we're interested in one. 18 00:00:59,358 --> 00:01:03,601 And try to hypothesize that we think that the average 19 00:01:03,955 --> 00:01:07,188 petal length for iris flowers 20 00:01:08,350 --> 00:01:11,330 is four centimeters. So, 21 00:01:12,543 --> 00:01:15,523 our null hypothesis would be 22 00:01:16,938 --> 00:01:19,514 that average, 23 00:01:19,514 --> 00:01:22,798 petal length is equal to four centimeters. 24 00:01:23,101 --> 00:01:27,092 And our alternative will be average 25 00:01:27,294 --> 00:01:31,487 petal length is not equal to four centimeters. 26 00:01:32,245 --> 00:01:33,962 Okay. 27 00:01:33,962 --> 00:01:35,630 The data we are going to use 28 00:01:35,630 --> 00:01:38,913 is the iris petal length data. 29 00:01:39,115 --> 00:01:41,136 So it's from the iris dataset. 30 00:01:41,136 --> 00:01:43,763 And this is the petal length and variable. 31 00:01:43,763 --> 00:01:47,400 Just to kind of remind us, it is just 150 32 00:01:47,400 --> 00:01:50,381 observations of different irises. 33 00:01:51,644 --> 00:01:54,624 To construct our test statistic 34 00:01:55,129 --> 00:01:58,716 we will first need an xbar value, 35 00:02:00,030 --> 00:02:02,202 which we can find by taking 36 00:02:02,202 --> 00:02:05,182 the mean of our sample. 37 00:02:06,344 --> 00:02:09,375 So the mean of the iris of petal length. 38 00:02:11,901 --> 00:02:14,781 Which will be 3.758. 39 00:02:14,781 --> 00:02:19,327 We also are going to need the hypothesized value 40 00:02:19,327 --> 00:02:24,076 that we are wanting to hypothesize, which is four centimeters. 41 00:02:24,076 --> 00:02:28,673 So I'm going to just call that mu because that's the parameter of interest. 42 00:02:28,673 --> 00:02:30,579 We're going to say it's equal to four. 43 00:02:31,906 --> 00:02:33,119 We also need to know 44 00:02:33,119 --> 00:02:36,756 the sample standard deviation, s. 45 00:02:37,110 --> 00:02:40,090 And so you can get that by doing this standard deviation 46 00:02:40,343 --> 00:02:43,323 of the variable. 47 00:02:43,677 --> 00:02:46,657 That value is 1.765. 48 00:02:47,112 --> 00:02:49,789 And then we also need to know the number of observations. 49 00:02:49,789 --> 00:02:53,023 So, n. So we will reduce the length function. 50 00:02:53,427 --> 00:02:57,115 And then I'll count how many observations are in your data set 51 00:02:57,115 --> 00:03:00,146 which is 50. 52 00:03:00,449 --> 00:03:02,419 Now once we have all of those 53 00:03:02,419 --> 00:03:06,107 individual pieces we can build the test statistic. 54 00:03:06,359 --> 00:03:10,805 Since we are doing a hypothesis test for a mean, we will be constructing 55 00:03:10,805 --> 00:03:15,907 what is known as like a t, a test statistic for a t-distribution. 56 00:03:16,816 --> 00:03:19,797 So I'm going to call it t-test stat. 57 00:03:21,060 --> 00:03:22,575 And how we create 58 00:03:22,575 --> 00:03:26,617 that is we do xbar minus mu in the numerator 59 00:03:27,476 --> 00:03:31,820 divided by I'm just gonna put this in parentheses 60 00:03:31,820 --> 00:03:36,821 as well. S divided by the square root of n. 61 00:03:37,529 --> 00:03:41,014 So thankfully we have all of these pieces already xbar, mu, 62 00:03:41,014 --> 00:03:44,803 s, and n. S, n, and xbar all come from the data. 63 00:03:45,409 --> 00:03:48,845 Mu is the value we specified in our null hypothesis. 64 00:03:49,350 --> 00:03:52,532 And this will compute our test statistic for us, 65 00:03:53,997 --> 00:03:54,907 which is 66 00:03:54,907 --> 00:04:00,918 -1.67897. So. 67 00:04:05,516 --> 00:04:07,435 Here we go. 68 00:04:07,435 --> 00:04:09,203 So our next step 69 00:04:09,203 --> 00:04:12,184 is to apply a decision rule. 70 00:04:16,175 --> 00:04:19,155 So we have two different ways we can do that. We'll- 71 00:04:19,762 --> 00:04:21,530 We will use a significance level 72 00:04:21,530 --> 00:04:25,015 or an alpha of 0.05. 73 00:04:25,369 --> 00:04:28,400 So I'm just going to go ahead and set that. 74 00:04:31,633 --> 00:04:32,694 And then if we want to 75 00:04:32,694 --> 00:04:35,675 calculate a rejection region, 76 00:04:35,686 --> 00:04:38,706 because there's two different kinds of decision rules we can do. 77 00:04:38,706 --> 00:04:40,120 Rejection region. 78 00:04:40,120 --> 00:04:43,101 We can find which critical value 79 00:04:43,202 --> 00:04:47,496 will give us a tail probability of 0.0 uh- 80 00:04:48,456 --> 00:04:50,880 Or since we're doing a two sided hypothesis test, 81 00:04:50,880 --> 00:04:53,911 we'll do our alpha divided by two. 82 00:04:54,065 --> 00:04:55,225 I'll kind of show you. 83 00:04:55,225 --> 00:04:58,256 So our rejection region 84 00:04:58,559 --> 00:05:01,893 is we're going to try, we're going to find the critical value 85 00:05:01,893 --> 00:05:07,198 that, fits the t-distribution, where the probability in the tail 86 00:05:08,612 --> 00:05:11,037 is equal to alpha over two. 87 00:05:11,037 --> 00:05:14,725 Because we're doing a two sided interval hypothesis test. 88 00:05:15,533 --> 00:05:19,878 Our degrees of freedom is needed for the t-test, which is n minus one. 89 00:05:20,837 --> 00:05:27,152 And since we are, our test statistic with a negative value, 90 00:05:27,405 --> 00:05:30,537 meaning that it's on the left side of the, 91 00:05:31,295 --> 00:05:34,225 of the mean on the curve, 92 00:05:34,225 --> 00:05:37,710 we will go ahead and say lower.tail equals true. 93 00:05:38,165 --> 00:05:43,116 Because we want the lower tailed like or the smaller the tail end probability. 94 00:05:43,469 --> 00:05:48,319 If this is a positive number 1.67 we would then do lower.tail 95 00:05:48,370 --> 00:05:51,906 equals false because we want the upper tail. 96 00:05:53,118 --> 00:05:56,048 We want kind of the extremes. 97 00:05:56,048 --> 00:05:59,938 So anything from where our test statistic is and more extreme. 98 00:06:00,999 --> 00:06:03,121 So what this will tell us 99 00:06:03,121 --> 00:06:06,152 is our, oh, alpha not found. 100 00:06:06,152 --> 00:06:08,627 I forgot to run that line. There we go. 101 00:06:10,650 --> 00:06:11,406 Okay. 102 00:06:11,406 --> 00:06:17,367 So our rejection value is one, -1.976. 103 00:06:18,074 --> 00:06:21,964 So what this is telling us is that if our test statistic 104 00:06:21,964 --> 00:06:26,611 is equal to -1.976 or less, 105 00:06:27,470 --> 00:06:32,876 or if it's greater than positive 1.976, 106 00:06:33,381 --> 00:06:36,412 then we will reject our null hypothesis. 107 00:06:39,443 --> 00:06:43,282 And in this case, since our test 108 00:06:43,282 --> 00:06:48,284 statistic is not in the extreme, it's actually greater than this value, 109 00:06:48,637 --> 00:06:51,618 we will fail to reject our null hypothesis. 110 00:06:51,618 --> 00:06:54,598 So this is telling us that, 111 00:06:55,558 --> 00:06:56,973 we will fail to reject 112 00:06:56,973 --> 00:07:00,812 our null, meaning that we do not have enough evidence to conclude 113 00:07:00,963 --> 00:07:04,803 that the average petal length is not equal to four centimeters. 114 00:07:06,015 --> 00:07:09,754 The other way you can apply a decision rule is with a p-value. 115 00:07:10,915 --> 00:07:13,340 And since we are doing a 116 00:07:13,340 --> 00:07:15,462 two sided hypothesis test, 117 00:07:15,462 --> 00:07:19,402 we will, can do two times whatever probability 118 00:07:19,402 --> 00:07:22,434 we get because we're going to be calculating it for one tail. 119 00:07:22,434 --> 00:07:25,439 But since we're doing two sided we'll just need to multiply it 120 00:07:26,020 --> 00:07:26,941 by two. 121 00:07:26,941 --> 00:07:29,809 And so what we're going to put in here is we're going to put 122 00:07:29,809 --> 00:07:33,749 in our test statistic that we get. 123 00:07:35,316 --> 00:07:38,296 The degrees of freedom again 124 00:07:38,296 --> 00:07:40,468 and again we're going to do lower.tail 125 00:07:40,468 --> 00:07:44,308 equals true because our original test statistic is negative. 126 00:07:44,308 --> 00:07:48,097 So we want a lower tail like the extreme value. 127 00:07:49,460 --> 00:07:53,335 And then we're going to multiply by two again because we are doing a two sided 128 00:07:53,704 --> 00:07:56,735 p-va- two sided hypothesis test. 129 00:07:57,139 --> 00:07:59,918 And then this is the value that we compare to 130 00:07:59,918 --> 00:08:02,898 our alpha, which is 0.05. 131 00:08:02,999 --> 00:08:06,485 So if our p-value is less than the alpha 132 00:08:06,738 --> 00:08:10,829 less than 0.05, we would reject the null hypothesis. 133 00:08:11,032 --> 00:08:14,416 In this case our p-value is greater than 0.05. 134 00:08:14,820 --> 00:08:18,357 So we would fail to reject our null hypothesis again as well. 135 00:08:19,316 --> 00:08:22,348 You should get the same conclusion. 136 00:08:23,004 --> 00:08:25,581 With either method, you should be 137 00:08:25,581 --> 00:08:29,117 coming to the same reject or fail to reject. 138 00:08:29,117 --> 00:08:32,198 You should not be getting different conclusions. 139 00:08:35,081 --> 00:08:36,491 So that's how you can kind of 140 00:08:36,491 --> 00:08:39,321 compute a hypothesis test by hand. 141 00:08:39,726 --> 00:08:43,767 But as always, usually in R there is an easier way to do it. 142 00:08:44,424 --> 00:08:47,253 So there is a function t.test 143 00:08:47,253 --> 00:08:50,233 which may be familiar from when we did. 144 00:08:50,435 --> 00:08:52,406 Confidence intervals for means. 145 00:08:52,406 --> 00:08:54,123 And this is actually you can 146 00:08:54,123 --> 00:08:57,104 do confidence intervals plus hypothesis testing in here. 147 00:08:57,710 --> 00:09:01,044 So we still are going to have the same null. 148 00:09:01,044 --> 00:09:04,075 And I turned it off hypotheses from up here. 149 00:09:04,530 --> 00:09:07,763 And so what we're going to do is we're going to just say t.test, 150 00:09:09,177 --> 00:09:12,208 give it the data that we are doing the t-test on, 151 00:09:13,370 --> 00:09:16,351 which is the petal length of iris flowers. 152 00:09:17,058 --> 00:09:20,089 We need to specify what our 153 00:09:20,190 --> 00:09:22,767 null hypothesis new value is. 154 00:09:22,767 --> 00:09:27,414 We're saying that we are hypothesizing that the true, average 155 00:09:27,414 --> 00:09:29,334 petal length is four. 156 00:09:29,334 --> 00:09:32,315 So we will say mu is equal to four. 157 00:09:32,567 --> 00:09:35,548 And then we also need to specify that our, 158 00:09:35,548 --> 00:09:39,235 our, that our alternative hypothesis is a 159 00:09:39,539 --> 00:09:42,519 two sided hypothesis test. 160 00:09:43,378 --> 00:09:45,747 Okay. 161 00:09:45,747 --> 00:09:47,672 And if we go ahead and run that. 162 00:09:47,672 --> 00:09:51,764 And notice it shows it is a one sample t-test which is perfect. 163 00:09:51,814 --> 00:09:53,936 We have one sample and a t-test. 164 00:09:53,936 --> 00:09:57,725 It gives us a t which is our test statistic 165 00:09:58,129 --> 00:10:00,857 which should match what we got up here. 166 00:10:00,857 --> 00:10:02,878 And it does. 167 00:10:02,878 --> 00:10:05,202 The degrees freedom is pretty easy. 168 00:10:05,202 --> 00:10:08,182 150 minus one. And then here's a p-value 169 00:10:08,283 --> 00:10:11,112 Same exact p-value we got here by doing a by hand. 170 00:10:12,963 --> 00:10:13,739 And then 171 00:10:13,739 --> 00:10:16,720 you can kind of see they have xbar right here. 172 00:10:17,477 --> 00:10:20,963 And then it also gives you that 95% confidence interval. 173 00:10:22,226 --> 00:10:24,398 So this is an, quick and easy way 174 00:10:24,398 --> 00:10:27,379 that you can compute a t-test for me. 175 00:10:28,541 --> 00:10:31,521 You can this is kind of showing you how to do it all by hand. 176 00:10:31,796 --> 00:10:35,217 And then this will show you kind of how to just do it in one simple step 177 00:10:35,217 --> 00:10:37,988 by computing a p-value for you. 178 00:10:39,099 --> 00:10:41,978 If you wanted to change what your, 179 00:10:41,978 --> 00:10:43,544 your null hypothesis was. 180 00:10:43,544 --> 00:10:48,041 So say, like you were testing, is the mean equal to two instead? 181 00:10:48,748 --> 00:10:50,213 You could totally do that. 182 00:10:50,213 --> 00:10:54,557 And then you can see that this p-value is way, way smaller. 183 00:10:55,669 --> 00:10:59,407 Or if you wanted to change your alternative. 184 00:10:59,559 --> 00:11:03,651 So it's not that it's just not equal to four and it's, you know, 185 00:11:03,651 --> 00:11:07,389 maybe less or greater than. So 186 00:11:08,209 --> 00:11:09,511 you could do it like this. 187 00:11:09,511 --> 00:11:12,491 You can do less or 188 00:11:12,895 --> 00:11:15,876 greater and that'll tell you, 189 00:11:17,694 --> 00:11:19,260 which, that'll 190 00:11:19,260 --> 00:11:22,999 change the output of your hypothesis test, 191 00:11:22,999 --> 00:11:26,181 kind of depending on if you're doing a one sided or two sided test.