1 00:00:01,200 --> 00:00:02,280 [Instructor] Now I'm going to show you 2 00:00:02,280 --> 00:00:06,570 how to do both a binary and an ordinal regression. 3 00:00:06,570 --> 00:00:08,760 So for the binary, 4 00:00:08,760 --> 00:00:12,370 our dependent variable is this one 5 00:00:13,290 --> 00:00:16,500 where we ask them is the environment 6 00:00:16,500 --> 00:00:19,290 or how important are environmental impacts 7 00:00:19,290 --> 00:00:22,530 when you buy food on a five-point scale?" 8 00:00:22,530 --> 00:00:24,360 And if they said four or five, 9 00:00:24,360 --> 00:00:26,190 it's important or very, 10 00:00:26,190 --> 00:00:28,110 we code it as one. 11 00:00:28,110 --> 00:00:30,900 If they said one, two, or three, 12 00:00:30,900 --> 00:00:33,150 they were coded as zero. 13 00:00:33,150 --> 00:00:37,980 And these are actually from 2013 Vermonter Poll data. 14 00:00:37,980 --> 00:00:40,200 For this exercise, we're just gonna do 15 00:00:40,200 --> 00:00:45,200 the demographic variables such as how much education, 16 00:00:45,780 --> 00:00:46,980 where they live, 17 00:00:46,980 --> 00:00:49,438 whether they own their home, race, 18 00:00:49,438 --> 00:00:53,823 political affiliation, gender, 19 00:00:55,440 --> 00:01:00,440 and then their income, household size, 20 00:01:00,660 --> 00:01:05,160 how many minors are in their household and their age. 21 00:01:05,160 --> 00:01:10,160 So to do this, we go to the analyze menu and regression, 22 00:01:10,170 --> 00:01:12,723 and we go to binary logistic. 23 00:01:13,980 --> 00:01:17,730 And I'm gonna reset this so you can see what I do. 24 00:01:17,730 --> 00:01:22,450 So again, our dependent is this environment 25 00:01:25,185 --> 00:01:26,820 with very important, 26 00:01:26,820 --> 00:01:31,820 and we're just gonna do all of these demographic ones 27 00:01:32,100 --> 00:01:35,973 like this and like this. 28 00:01:46,470 --> 00:01:48,303 And then once that's done, 29 00:01:49,470 --> 00:01:51,663 you can just hit okay. 30 00:01:54,870 --> 00:01:59,870 So you can see here that 31 00:02:05,280 --> 00:02:10,170 this is the equivalent of sort of like the F stat. 32 00:02:10,170 --> 00:02:13,290 So this is minus two log likelihood, 33 00:02:13,290 --> 00:02:18,290 and we'll have a look at this when we do a nested model, 34 00:02:19,140 --> 00:02:23,400 but this is what we wanna see here, 35 00:02:23,400 --> 00:02:27,213 mostly focusing on beta and the significance. 36 00:02:28,170 --> 00:02:31,050 So the only things that are significant 37 00:02:31,050 --> 00:02:35,140 are those who live in a single family dwelling 38 00:02:36,020 --> 00:02:37,027 have a negative sign 39 00:02:40,530 --> 00:02:42,956 which means they're more likely, 40 00:02:42,956 --> 00:02:44,160 less likely to say that. 41 00:02:44,160 --> 00:02:49,160 Those that own their own home have a positive impact. 42 00:02:50,670 --> 00:02:55,670 And last female again has a positive impact. 43 00:02:59,280 --> 00:03:02,850 This wall is a lot like a T stat. 44 00:03:02,850 --> 00:03:05,940 So you can see that a big wall 45 00:03:05,940 --> 00:03:08,790 leads to a small significance here. 46 00:03:08,790 --> 00:03:13,410 A small wall leads to a big significance. 47 00:03:13,410 --> 00:03:18,410 So one of the things that you may choose to do 48 00:03:19,590 --> 00:03:24,590 is to do a nested model. 49 00:03:25,860 --> 00:03:27,810 Only those that are significant 50 00:03:27,810 --> 00:03:32,070 or maybe only those where the wall is greater than one. 51 00:03:32,070 --> 00:03:35,173 So here we could do that. 52 00:03:35,173 --> 00:03:39,540 So that would be just rural, single family dwelling, 53 00:03:39,540 --> 00:03:42,630 own home, Democrat and female. 54 00:03:42,630 --> 00:03:47,630 So, and note this number here, the log likelihood 826 55 00:03:51,030 --> 00:03:54,480 because that's what we're gonna be comparing. 56 00:03:54,480 --> 00:03:58,680 So now we're gonna run this same thing, 57 00:03:58,680 --> 00:04:02,085 another binary logistic, 58 00:04:02,085 --> 00:04:05,703 and we're gonna get rid of all of these. 59 00:04:08,017 --> 00:04:09,503 Make sure that we are gonna keep, 60 00:04:10,440 --> 00:04:11,973 so these three. 61 00:04:12,840 --> 00:04:17,310 White goes away and Democrat stays 62 00:04:21,900 --> 00:04:26,900 and female stays and all of the others go back. 63 00:04:27,000 --> 00:04:30,240 So we're left with one, two, three, four, five. 64 00:04:30,240 --> 00:04:32,370 Let's just make sure that's right. 65 00:04:32,370 --> 00:04:34,803 One, two, three, four, five is right. 66 00:04:35,640 --> 00:04:37,623 And we would run this. 67 00:04:42,600 --> 00:04:45,780 And we again look at this log likelihood. 68 00:04:45,780 --> 00:04:49,830 So here we see that there's a very big change, 69 00:04:49,830 --> 00:04:53,550 that it's gone from 800 and something to 900 and something 70 00:04:53,550 --> 00:04:57,900 which makes us say that those variables matter. 71 00:04:57,900 --> 00:05:01,200 There was a big change when we excluded them, 72 00:05:01,200 --> 00:05:04,470 and we're much more likely to keep the full model 73 00:05:04,470 --> 00:05:08,010 and not use our restricted model. 74 00:05:08,010 --> 00:05:11,190 But if you wanna see, it didn't really change much 75 00:05:11,190 --> 00:05:14,790 with the significance female. 76 00:05:14,790 --> 00:05:18,300 It is still sort of the biggest walled 77 00:05:18,300 --> 00:05:21,240 and the most sign significant by far. 78 00:05:21,240 --> 00:05:25,830 But we would use our full model that we see here. 79 00:05:25,830 --> 00:05:29,250 And these are the factors that drive 80 00:05:29,250 --> 00:05:34,250 whether or not someone found the environment very important 81 00:05:35,610 --> 00:05:38,370 when they do, when they buy food. 82 00:05:38,370 --> 00:05:42,690 So one more thing that I would like to do now 83 00:05:42,690 --> 00:05:44,910 is to do an ordinal. 84 00:05:44,910 --> 00:05:49,910 So we just do regression and ordinal 85 00:05:51,870 --> 00:05:55,380 And I'll reset it just so you know what we're doing here. 86 00:05:55,380 --> 00:05:59,550 So the dependent is this ordinal variable 87 00:05:59,550 --> 00:06:04,550 of how often do you eat as many fruits and vegetables 88 00:06:04,710 --> 00:06:09,710 as you should according to public health guidelines. 89 00:06:10,890 --> 00:06:14,073 And then I believe this was on a four point scale, 90 00:06:15,360 --> 00:06:19,230 sort of one being, you know, hardly ever, 91 00:06:19,230 --> 00:06:22,590 four being almost always or something along those lines. 92 00:06:22,590 --> 00:06:23,880 So for this, again, 93 00:06:23,880 --> 00:06:27,183 we're just gonna look at all the demographics. 94 00:06:28,457 --> 00:06:31,750 So we're gonna say, and we're gonna code them as covariates 95 00:06:33,120 --> 00:06:36,360 because sometimes when you do 'em as factors 96 00:06:36,360 --> 00:06:41,360 which are the binary ones, they flip the the signs. 97 00:06:41,670 --> 00:06:44,703 So I suggest always doing them as covariates. 98 00:06:45,540 --> 00:06:50,540 And then we can look at these four demographics as well. 99 00:06:52,110 --> 00:06:54,660 For the actual article that we wrote, 100 00:06:54,660 --> 00:06:59,660 we included these attitudinal and behavioral ones too. 101 00:07:02,670 --> 00:07:07,050 But for this one we're just gonna do these demographics. 102 00:07:07,050 --> 00:07:11,760 So we hit okay, and then we look at what we get. 103 00:07:11,760 --> 00:07:16,760 And here again, we focus on what's significant. 104 00:07:19,680 --> 00:07:24,103 So looks like education is significant, 105 00:07:25,200 --> 00:07:28,770 being an independent voter is barely. 106 00:07:28,770 --> 00:07:32,503 Being female and having higher income, 107 00:07:35,340 --> 00:07:39,180 so all of these have a positive sign. 108 00:07:39,180 --> 00:07:43,473 So we can include that education, 109 00:07:45,750 --> 00:07:49,620 being an independent voter, being female 110 00:07:49,620 --> 00:07:53,400 and having a higher income on average 111 00:07:53,400 --> 00:07:57,940 all lead to somebody being more likely 112 00:07:59,520 --> 00:08:00,771 to eat fruits and vegetables 113 00:08:00,771 --> 00:08:05,771 at the rate at which they should. 114 00:08:06,240 --> 00:08:09,600 And again, I was right, this is a four point scale. 115 00:08:09,600 --> 00:08:14,600 So the last thing is that I can show you the citation 116 00:08:15,330 --> 00:08:17,943 for the actual article that we did, 117 00:08:20,880 --> 00:08:22,650 and we're gonna look at that. 118 00:08:22,650 --> 00:08:27,650 And you can see that I published with Bernice Garnett, 119 00:08:28,770 --> 00:08:31,410 the fellow UVM faculty. 120 00:08:31,410 --> 00:08:36,410 Here's the citation and you can read it, 121 00:08:36,870 --> 00:08:41,870 and you can see that in our analysis we also included 122 00:08:47,760 --> 00:08:52,760 all these attitudinal and behavioral ones too. 123 00:08:52,992 --> 00:08:57,992 So that is how you do a binary and ordinal regression 124 00:09:00,120 --> 00:09:02,283 in SPSS.