302 email

> Hi, I have printed out the assignment 1 from the web! I am just > curious when you asked us to give the definition of something, what kind > of things do you expect us to write? > > For example, 1a) Give the definition of E[X] = u(mean)x > > Definition: The expected value of a random varibale X (x1, x2, x3, > ......) is the mean(u) of x. The answer you gave would get zero marks. Full marks for something like equation A.14 on Page 665 of the text. You should read this appendix of review material. I should have suggested this in the first lecture. > I want to know: how long of each quiz and > how many questions do we have in the quiz? Also, do we have past quiz on the > website? I mean: do we have practice quiz? How should I know your format > of the quiz? I plan about 20 to 25 minutes. Maybe 2 questions on average. The format of the quiz questions will be similar to the format of the questions on the asssignments. But quizzes and tests from past years are a good idea. I will start getting this material on the Web. > I attended the STA 302 lecture yesterday, however, I am not quite > clear about the CQuest. Is it a software? Cquest is a computer facility. It is in room 2105. Stop by there and take a look. > How can we access > the software/program at home? If you are a computer person, the following might be enough. The telnet address is server.sid.cquest.toronto.edu, but the people who run cquest think telnet is not secure enough. Instead, they use ssh (secure shell). See http://www.cquest.utoronto.ca/infodocs/cquest/droppingTelnet.html for information on how to obtain ssh that runs on a Windows machine. If you are not particularly a computer person, I will try to assemble some instructions in plain English. But if you can't make it work, you are still responsible for the computer part of the course; you may have to work in room 2105. > Do we have to use > any statistics software in order to do the assignments? In order to do some of them, yes. > What kind of assignment can we bring in the quiz? Only computer output from the computer part of the assignments. > How long will > the quiz be? Maybe 20 to 25 minutes. > For assignment 1, we don't have to use computer, > right? Right! By the way, if you can send me plain text email instead of html, I would appreciate it. > Hi, I'm in your STA302 summer class. I just want to suggest that it'd be > better if you would put down "last update..." on the course webpage. > Good suggestion, thanks. Done. Jerry > Hi, I am student of STA302F, during the beginning of the lecture you did > mention to bring printout of SAS DISPLAYS next class. Which files you are > referring to?! Should I print both intro.* file and reg1.* together and > bring to next time?! Thanks. > That would be fine, though it's a little more than you need. As the Web page indicates, you should print and bring * intro.sas * MAYBE intro.lst (this is optional) * reg1.sas * reg1.lst If, after the lecture, you feel you want copies of the other files, you can always print them then and include them with your class notes. > You mentioned in your email to another student that the definitions for some > of the review material was on p. 665 in text. What if your one of the > unlucky ones(like myself) who don't have a text? We can't get a copy of the > Appendix. I thought the answer the student gave you(which you said would > get him a zero), was alright. So now I don't know what is good enough. You have a good point about the text. I sent that answer assuming everyone had a copy. Well, note that the question said the random variable involved was CONTINUOUS. You need to get your textbook from the Stat course that was prerequisite to this one, and look up the definition of the expected value of a continuous random variable. Look in the index. The answer is expressed in terms of an integral. I suppose it is possible that if you are coming from ECO220Y, your text may not have defined the expected value of a continuous random variable. But I'll bet it did. The purpose of the review part of the assignment was to get you to review material you may have forgotten. Naturally, this includes looking at textbooks from courses you have taken. > In the assignment 1 question 1 a) , it ask for give the definition of > E[X] = mu . > In the website E-mail section, you answered that "Full marks for > something like equations A.14 on page. 665 of the text". I am one of > the student who still don't have the textbook. I already got a copy of > Ch .1 & 2 from SS library, but still don't have page.665 . So can you > tell me what equation appeared at page. 665?! Integral of y f(y) dy > > Also , can you give me a hint of working on the proof of Cov(X, Y) = > E(XY) - E(X)E(Y) ? You need to start with the definition Cov(X,Y) = E[(X-E(X))(Y-E(Y))], expand the product and distribute the expected value sign. Don't forget that E(X) is a constant that could be written mu_x and E(Y) is a constant that could be written mu_y. > > Is Var(aX+bY) = Var(aX) + Var(bY) - Cov(X, Y) ?? No. a^2 Var(X) + b^2 Var(Y) + 2 ab Cov(X, Y) > Thanks for posting the old quizzes, tests and exams, is it possible to post > the answers as well? No, I'm sorry, I have no idea where the old answer keys might be, or even if they still exist. > Since this course is quite intensive, I always > encounter some difficulties when doing the past paper or problems in the > books if there's no answer because I don't know whether my answer/ my > approach is right or not. Also, for the problems in our text, is there any > way for us to find out the solution? For some of the text, there's answer > for even #, but this text doesn't have any solution. > I agree it would be good, but again I don't have an answer key either and it would take too much time to make one up and get it distributed properly to the class. However, you are welcome to come to office hours and we can discuss your approach to the problems. > I am a student of STA 302 summer(2001), I know u have already posted the > list of some login names, but I can't find my name on it. Please help check > for me, thank you!! I can't find your name on the list either. It may be that you registered a little later than other students in the class and therefore your computer account will be created a little later too. In that case, the best thing would be to check the hard copy list outside the stat aid centre (SS 2133) on Tuesday or Wednesday and see if your name is there. If it is not, I will enquire. But if you registered for the course, say, more than two weeks ago, then I should start asking questions right away. When did you register? > > and also, can u tell me time & location for TA, thanks a lot!! > There is no tutorial, and no set location. There is a marker, a graduate student named Mohammad Dolatabadi. His email is mohammadd@utstat.toronto.edu, but he is not paid to answer questions about the course, just about the marking -- and not many of those questions, either. > can E(X+Y) written as double integral [ (x + y) f(x, y) dx dy ] ?? > Yes. > I am taking STA302 this summer. I > attempted the first assignment. Could you please see whether I am on the > right track or not? > > Answers: > Q1.5 No Correct, but why? > > Q1.7 a, No because we don't know the distribution of E(i)? Should we use CLT $ > not? Right. CLT does not apply; it's only one observation. > > Q1.7 b, We have to use normal (X- mu)/ (sigma/root(n)). Is n=1? Yes. > > Q1.10 Yes Nope. The data are a snapshot at one time point. The older programmers probably have been there a long time. They likely started at low salaries by modern standards, and their raises have been overtaken by entering salaries of younger (and probably more valuable) employees. > > Q1.29 "What is the implication of regression function is beta(subscript o)=0" > What is theh above question asking? Is the answer: E(Y)=0 when X=0 the right > answer to this question? Yes. What would the graph look like? > I'm a student in your STA 302/1001 course. i have a couple of questions > regarding assn # 1 which i understand is very important for Quiz #1. My > first Q) pertains to Q. 6, on matrices. What do you mean by a counter > example? It is claimed that the sum of two odd numbers is always odd. The claim can be disproved by noting that 3+3=6, which is even. Similarly, for example, the claim that AB=BA is false in general for matrices. Provide a counterexample by giving two specific matrices (I's suggest both 2 by 2) for which AB does not equal BA. > Secondly, It's been ages since i took a course in vectors and > matrices and I'm wondering what the prime (') in parts d) and e) stand for. > Is it equivalent to the transpose of a matrix? Yes. > Lastly, where do i find the > solutions to these problems? I'm sorry; they are not available. > Surely, i would need to refer to a stats text > book as your class notes are not sufficient for the review part. Right? Right. > Can you give me some hint on how to show mu-zero is outside the > (1-alpha)100% Confidence Interval for mu, by using 3a) ?! Write down the decision rule for the t-test. Assume t* exceeds the critical value on the upper tail. Do some algebra. Now assume t* is less than the critical value on the lower tail. Do some algebra. It's an if and only if. Do you have enough to go in the other direction? > I have some more questions regarding Assn 1. Q 1.27 from problem set of > chapter refers to first order regression model shown in Chapter 1 of the > text book which i don't have. What is this model? The first model given in lecture notes. No normality assumption. >

1.For discrete iid when given a PrMassF > p(x;theta)

MLE can be found by multiplying 2 PMF together, > taking the log/ln(?) and then taking the derivative with respect to > theta.

Two? There would be n. Otherwise I agree. Set the derivative to zero and solve. >

Question: Is the answer always the same in general:

(Sum of Xi from i=1 until n) Sigma Xn/n

Does this make sense?

It makes sense, but the answer is no. > I'm happy that chpt 1&2 are now available but i'll have a problem gettting > to the library. I work full time in the day time and i would have to take > some time off during work hours to go to downtown given that the library is > only open till 5 p.m. Any possibility you can arrange to have chpt 1&2 > faxed to me? Can the textbook be found at a library in erindale? No, but the book does seem to be at Gerstein library on short-term reserve. I am trying to get some of the photocopies moved to Robarts, but in the meantime there is Gerstein. > > For the Sta 302, is it possible for you to put a copy of solution manuel > in the library so that we can photocopy it ourselves? Yes, it's now on short term reserve at Gerstein. > Also there's a disc at the back of the text book, what > is it for? Data used in problems and examples in the text. > I am using a used book, there's no disc when I bought the text. > Is it possible for me to get a copy of the disc/ or the data from the > disc? Whenever I assign something using these data, I will also post the data on the Web, which is probably illegal, but who cares? > i was wondering if you will go over all the quiz > problems in the lecture. if not, can you possibly > post the answers on the web or on your door if that > makes it easier for you. If Mohammad remembers to give me back my solutions (as I asked him to) I will post photocopies. Later, I will put it on the Web as well, but right now my scanner is at home and the solutions are at school -- so we'll start with photocopies.

> do we need to print out any SAS output from the > CQUEST acct to bring to Quiz2? or the asst 2 > sheet? thank you! No, I'll provide everything. > For assignment 2, do we need to use any sta software to do it? No. But you may be given some SAS output similar to what is on the assignment sheet, and you may be asked questions about it. But for Assignment 2, you do not have to run any computer jobs or produce any output yourself. > Also, I am > not quite sure about the STA software, is SAS the program that we are going > to use for our course? Yes. > Do we have to download the sta program? I am still > a bit confusing about the sta software, where can I find more information > about this? I don't mean to be rude here, but if you have questions of this nature, you need to ask them in class when I am trying to tell you about the computer stuff. This way, everyone can benefit from the answer at the same time. Of course, you are welcome to come ask me during office hours too, but PLEASE, don't just sit there politely in class while I'm trying to explain something you're not familiar with. You have some responsibility for your education too, and I am VERY willing to be guided by students' questions. Please speak up! (And please don't take this lecture personally. It applies to lots of good students.) > I cannot read the 'Raw Data file'. What should I do? I assume you mean dwaine.dat. Well, I can read it, using Netscape over a modem from home. But it's not important. The data are in the book, and anyway all you are asked to bring to class are the command file and list file. > After I downloaded the file from the web site, I can't open the .dat > file. What program should I use to open the .dat files?????? It's a plain text file. All the files having to do with SAS are plain text files. Notepad or ANY word processing software should be able to read it. This may be a more helpful answer to the previous question too. >

Hi, you said 2.8 > won't be on quiz 2, so will 2.9-2.11 on quiz 2 as well? Yes, potentially. For example, you need Sect 2.9 to answer problem 2.23, though I also covered this material in lecture. > Also is > there any questions base on the matrix you taught last wednesday and > which chapter is that?

That stuff will turn up on Quiz 3. It's in Chapter 5. > I have a few question which I am not sure about my answer. would you please > confirm them. > > Q 2.3) We can't conclude this because of the other reasons (regression and > causality) Look at the p-value. What does it tell you? > > about Questions 2.9 and 2.11, would you please give me hints > Q2.9 hint: How many potential Xh values are there? Q2.11 hint. One of these means is an arithmetic mean, and one is an expected value, that is, a population mean. Do you know which is which? > Question 2.1b > What is this question really asking us to discuss? The meaning, if any, of the intercept in this situation. > > Question 2.3 > What does the two-sided p-value for estimated slope tell us?? > I am really confused with the two-sided p-value stuff. I'm sorry, I cannot type you a lecture on this topic. Please see the textbook of the statistics course you took as a prerequisite to this one, and read about p-values. It is a central topic in hypothesis testing, and will be thoroughly discussed regardless of the course you took or the text that was used. > Question 2.17 > Is the answer to ques. 'was the alpha level used by the analyst greater than > or smaller than 0.033?' ---> since it was a rejected conclusion (beta 1 not > equal to zero), then the alpha level was greater than 0.033 (pvalue) Yes. > > For the question 'if the alpha level had been 0.01, what would have been the > appropriate conclusion?' is the answer ---> since the p-value in this case > is greater than alpha level, we would accept and thus a conclusion of H0: > beta1 = 0 would be reached Just say H0 was not rejected, and there was not enough evidence to conclude that a linear relationship exists between X and Y. In this course, we will never accept the null hypothesis, just sometimes fail to reject it. There was a bitter dispute about this in the early history of Statistics, and we are siding with Mr. Fisher, after whom the F-test was named. You can not lose any marks because of this subtle distinction, however. > > Question 1.19b > How do we create the scatterplot if we are not given all the points? You have a good point. Sorry about that. You are NOT expected to use the data disk that came with the text, for this question. > > Question 2.23a > Arent we already given the ANOVA table on the assignment sheet? Well yes. I hope you agree that it makes the problem easier. Lots of the answers here are directly on the printout. You just have to know how to find and identify them. > > Question 2.23b > Isnt F* estimated by MSR and by MSE? See the text. I also covered this (fast) in lecture. > However, what is the condition that > MSR and MSE estimate the same quantity? > When H0:beta1=0 is true. See the text. > Hi, I just got a passing grade on quiz 1 and I don't think I did > really good in quiz 2. I am just wondering is there any possbility that > the course mark will bell up at the end of the term. I bump people up ONLY when they are right at a letter grade boundary; I don't bell. Whether everybody gets A+ or everybody fails, I just turn in the marks the students have earned. > Since as I see the > mark list on the web page, many people's quiz one marks are quite low. In my opinion, there is (or was) a distinct cluster of people whose background knowledge is not enough to succeed in this course at the accelerated summer pace. Since you passed quiz one, I would NOT necessarily put you in this group. > Do you suggest me to drop this course while it is not too late? If you had gotten a zero or one on quiz one, I would say yes. But in your particular case I just don't know. I will say that quiz two will probably be the easiest one. The decision is up to you. > > I have a question regarding to SAS print out. I tried to print my log and > list files but I wasn't able to do it.I save my log file as a test.log file > and my list file as a test.txt and for printing command I used (lpr > test.log;) and (lpr test.text;) BUT I didn't get any printout. I typed them > in program editor window.Would you please tell me where my mistake is. > The lpr command is in reponse to the unix prompt. As I said in class, I recommend AGAINST doing anything in the SAS program editor. It's not wrong, it's just less convenient than the way I'm trying to show you. > > Q1) In the Problem 1.21, part (b), is the "point of the expected number" the > same as the Yhat ? Point ESTIMATE. Yes. > Q2) In the SAS handout no.4, Ch6b.sas, what is the meaning of "proc glm"? It's another linear regression program withinn SAS. I hope this was covered adequately in class. > So far I have typed someting in > a window called "SAS:Program editor". Is this the right one? As I said in class, I recommend AGAINST doing anything in the SAS program editor. It's not wrong, it's just less convenient than the way I'm trying to show you. > How come the > log file didn't show what I've typed in so far? Also, Which window is > suppose to containt he *.lst file as I need to print them out. I opened the > SAS program through the statistics homepage where it says "Softwares:SAS" Please try to do it the way I'm describing in class, and if I'm not being clear please speak up! What you're doing here is to follow your nose the way you might in Windows or on a Mac, and even though the person who set up that link on the Stat homepage may have meant well, it is the road to trouble. > Please let me know if I am using the SAS I'm suppose to be using. Yes it's the right program, but there is more than one way to use it, and I am not teaching or supporting the approach you're taking (which, by the way, IS the approach recommended by the SAS company). The approach I'm suggesting is use a text editor to create a command file and maybe a data file, and then run SAS from the command line. Suppose your comand file is called hw3.sas. You would type sas hw3 in response to the uniz prompt, and then take a look at hw3.log and hw3.lst using the less command. > > I have a question regarding to Question 6.4. Why don't we attach a sign to > the correlation cefficient of multiple correlation? > Well, suppose p=6, two of the b coefficents are positive and 3 are negative. What sign would you attach? Maybe one of the positive coefficents is very large while all the negative coefficents are all very small in absolute value ... > when we're asked to test physicians and hospital beds while keeping > everything else constant, does that mean were testing if: > > 1) beta(physicians)=beta(hospital beds)=0 or are we testing > > 2) beta(physicians)=beta(hospital beds) ? Well, the second test would be asking if increasing the number of physicians by one would have the same effect on expected number of crimes as increasing the number of hospital beds by one. This would really be weird. In number 1, the null hypothesis says neither variable is related to E[Y] -- more of a reasonable possibility. > For Question 8 on the exam preparation SAS assignment I got and F of 28.128. > The line of code I used was > > model InfRisk = LStay NumNurse RCR RCXray / ss1; > > I can't figure out what the problem is that I don't get what you had > suggested 19.25. > > Thanks. > You are reading the F statistic automatically generated by SAS. It's for a different null hypothesis than the one you want. You need a test statement in proc reg. > what section of the textbook we need to read after quiz#5? > is that only SAS assignment we need to do, nothing from the textbook?. Section 8.3 and 9.1-9.4 should help you understand what you're doing with that SAS assignment. But new material that is not mentioned in the assignment will not be on the final. > Hi, I just checked your suggested solutions of quiz 5 on the STA302 > course web site. I have a question about 2c's solution. > > In Q2c, my answer was Ho: B2=B3=B4=B5=B6=B7=0. And my C matrix was > a 6x8 matrix which is shown below: > > c= 0 0 1 -1 0 0 0 0 > 0 0 0 1 -1 0 0 0 > 0 0 0 0 1 -1 0 0 > 0 0 0 0 0 1 -1 0 > 0 0 0 0 0 0 1 -1 > 0 0 0 0 0 0 0 1 > > However, your answer is: > > c = 0 0 1 0 0 0 0 0 > 0 0 0 1 0 0 0 0 > 0 0 0 0 1 0 0 0 > 0 0 0 0 0 1 0 0 > 0 0 0 0 0 0 1 0 > 0 0 0 0 0 0 0 1 > > 1.) I would like to know that if my answer for 2c is correct. If it's not, > why? and when should I use my method? Our matrices are row equivalent. They are both right; they make the same statement about the betas. There are always infinitely many correct and equivalent ways to specify a null hypothesis. > > 2.) In Q1, I just wrote down "the confidence interval is 77.275 plus or > minors 4.774" but not (72.501, 82.049). Will I get full marks for Q1? I assume so. If not, we will fix it. > > 3.) When will we get back the last quiz? Will we get it back before the > final exam? If yes, how? Office hours Monday, 4-6 or Wednesday 3-5. > on the 3rd paragraph of the final assignment, it asks us to "Make dummy > variables for region", but do we also need to create dummy variables for > Medical School Affiliation? > If you write down E[Y] for each of the two medical school affiliations, you'll see that the same information is available whether you do or not. This is a good exercise. > 1) at first it asks us to state the meaning of b0, is it the b0 while not > treating medical school as a dummy variable? Yes, leave it alone for this one. > 2) the same question asks us to re-express it as 0-1 variable, that means > we're going to do another proc reg for medical school while treating it as a > dummy? Not needed, though you can if you want. The question just asks you to say what b0 WOULD mean. > 3) and also the test statement for this question would be according to > medical school as a dummy? I think it may be valuable for you to do it both ways after all. Does it matter? > question 8: > i have read the Email question and answer page, i don't get 19.25 either. my > command as follows: > > proc reg; > model risk = stay nurses culture chest / ss1; > risk4: test stay=nurses=culture=chest=0; > > the F value from the test statement result was 28.128. is there any problem > with the above command? > Yes. Please read the question again, carefully. "They" refers to the two ratio variables, right? What else could it mean? > When I read the notes from June 18 ( Mon), I found > that there are sections about Outlier, Deleted > Studentized Residual, DFFITS and Cook's distance. I > want to make sure that are those materials will on the > exams as well? If so, can you specify the > corresponding sections in the Textbook? Thank You . I intended to put that stuff in the last assignment but I ran out of time. I think I've got deleted Studentized residuals with a Bonferroni correction on the assignment, so they MAY be on the final. Things I did not put on the assignment are not allowed. So I guess you could get by with Sections 8.3 and 9.2. Plus whereever Bonferroni is to be found (see index). > > 2. For Q 3, does b-not mean: if there is no Medical School Affiliation then > InfRisk would be 5.964277? No. Hint: how is medical school affiliation coded? > is it correct to conclude, Hospitals WITH a medical school affiliation tend > to have lower infection risk.? Again I ask you to look carefully at how medical school affiliation is coded. What does 1 mean? What does 2 mean? > > 4. for Q4, what value tells proportion of remaining variation in DV after > controlling for one of the IV...I know that R-square explains the variation > in DV but does it not explain the variation using ALL IV in the model? Yes, so after you fit a reduced model, some of the variation is left unexplained. Then you fit a full model. How much MORE is explained. Express this as a proportion of the variation that was left unexplained by the reduced model. > > 5. What does Adj r-square mean? See Equation (8.4) on P. 339, but it's not on the final. > > 6. For Q8, we use a model with the 4 vars: length of stay, num of nurses > and the 2 ration vars, but what are we supposed to test? The two ratio variables. > > is there > another part to assignment 6 onthe net that I dont see aside from the > sas stuff? > No, nothing extra that's not on the assignment. > I was a bit > confused. In question 2, the independant vbl is nurses. Does that mean > that the model is: > > model x11= x4; because it is simple regression? Or are there more > variables in the model? That's all, but may I strongly suggest different variable names? You are going to have to read your printouts under time pressure, and you should try to make it easier on yourself. > > In question 4, Do you fit a model with 2 IV's like the following: > model x10 x11=x2 x3 x4 x5 x6 x7 x8 x9 x12; > or is it: > model x10= x2 x3 x4 x5 x6 x7 x8 x9 x11 x12; > model x11= x2 x3 x4 x5 x6 x7 x8 x9 x10 x12; Two independent variables means two predictors to the right of the equals sign. There s always just one dependent variable - infection risk. > > In question 5, HOW DO YOU SET IT UP. If all the variables are independant, > what is dependant? The ones you are excluding (x5 x6)? Again, there is always just one dependent variable - infection risk. > > In question 4, when you ask for the proportion of the remaining > variation in infection risk that is explained by number of nurses > (controlling for average daily census), I'm not sure which expression > you're asking for. Is it [SSR(F)-SSR(R)]/SSTO or [SSR(F)-SSR(R)]/SSE(R) > ? > The latter. I would call the first one the proportion of additional variation, or something like that. > Also, in question 8 when you ask, "Is it significant at alpha=0.05?" are > you asking whether or not the two ratio variables are significant? I > mean, should I be checking if F=19.25>F(.95,2,108)? > Yes. > i tried question 8 but got a value of ~28 for F not 19.25. Any suggestio$ > as to what may have gone wrong. I've pasted the list file below, where > var4=infection rate, var 2= length of stay, var 11 = # nurses and var 5 and v$ > 6 coorespond to the two ratio terms > The right test has 2 df, not 4. See the email questions file for more detail. Also, please use more informative variable names, for your own good. The email questions file has some discussion of this, too. > Hi, I just want to make sure that , is the final exam open book and notes? Open book, CLOSED notes. >

Hi, could you give me a hint on how I would do part 3 - > estimate the difference in expected infection risk between hospitals > with and without a medical school affiliation. Do I need to create > a dummy variable for this? Figure out the expected value of infection risk for hospitals with a medical school affiliation. Then figure out the expected value of infection risk for hospitals without a medical school affiliation. Do this first without any dummy variable coding. Then decide if you need dummy variable coding or not. > Hi, the exam will cover ch2,5,6,7,11, and 9.2 only? > is 8.3 in the exam? See assignments 2 through 5 for exact coverage in the textbook. Beyond what is in assignments 2 through 5, you are responsible for Sections 8.3 and 9.2. See email questions and answers for more detail. > and also, can i bring the formula sheet that your > posted on web to the exam? No, but you won't need it anyway. All that material is in the text, and it's open book. What's NOT in the text is the multivariate normal stuff. That should be in your head, or if need be you can write notes in the margins of your text.

Quiz 2 Stuff