Leading On-Line Chat Sessions Space For Your Notes Purpose of this segment: In this segment we will discuss the idea of synchronous chat as well as the alternative asynchronous threaded discussion approach. We will consider the effectiveness and resource intensiveness of these two approaches in technology intense subjects as well as "softer" subjects. We will then look at ways to inspire the students to participate in the chats and how an instructor can be a part of the chat without turning it into a "mini lecture." And then we will look at the logistics of how to be organize for an effective chat by having the topics handy,Insight havingInfrastructure common responses to students’ questions and how to Tool archive the session. Space For Your Notes Leading On-Line Chat Sessions Professor Thomas B. Barker CI01 Insight Infrastructure Tool Space For Your Notes Some Approaches • Synchronous Chat • Asynchronous Discussion • Do nothing CI02 Insight Infrastructure Tool Page 1 Leading On-Line Chat Sessions Space For Your Notes Resources Synchronous Chat At least 1 hour per week but up to 2 hours for both the faculty and the students CI03 Insight Infrastructure Tool Space For Your Notes Effectiveness Synchronous Chat “The chat really brought the ideas in the course together” CI04 Insight Infrastructure Tool Space For Your Notes Resources Threaded Discussion About 30 minutes per week for the students (some will spend longer). Many hours for the faculty. CI05 Insight Infrastructure Tool Page 2 Leading On-Line Chat Sessions Space For Your Notes Effectiveness “..the threaded discussion was a complete waste of time as far as I was concerned..” “I spend as little time on the discission as possible. Just enough to show I was CI06 there and get my grade.” Insight Infrastructure Tool Space For Your Notes Should we be Attempting to Emulate the Classroom? LDTATNC* CI07 Insight Infrastructure Tool Space For Your Notes Chat Approaches • Free for all (do nothing) • Assigned topics in list • Assigned topics in Video (with list) CI08 Insight Infrastructure Tool *Let's Discuss This At The Next Chat Page 3 Leading On-Line Chat Sessions Space For Your Notes Which of the above do you think works best? LDTATNC CI09 Insight Infrastructure Tool Space For Your Notes My Experience • Assigned topics in Video (with list) Let’s look at an example chat Chat Archive 801 Tuesday 7PM 9/12/2000 CI10 Insight Infrastructure Tool Space For Your Notes CI11 Insight Infrastructure Tool Page 4 Leading On-Line Chat Sessions Tom Barker has joined the chat. Joan Gregorski: That’s because we’re not in mexico :-) Terrance Antinora: Hola Mike Laurel Harrington has joined the chat. I'll put the imporMikel Hill: That’s about all the Spanish I know. tant remarks made Tom Barker: Greetings to all. I’m glad to see so by the leader in many of you here already. italics. I usually Tom Barker: Well it’s 7:00:15 let’s begin Laurel Harrington: Hi. Spanish? change the color Donna Perkins has joined the chat. to red in the arDonna Perkins has left the chat. chives. I'll bold the Allal Ahmamad has joined the chat. Tom Barker: Donna is trying to use the WEB FC chat topics. Mikel Hill: Looks like it’s not working Allal Ahmamad: hi everyone Tom Barker: nope Mikel Hill: Prof. Barker: what are we supposed to do with these chat sessions? Tom Barker: OK The idea is to watch the tape and you will notice that the natural questions that would happen in a live class just can’t happen. Tom Barker: I have suggested some topics that would have come up for discussion and mentioned them in the tape …and listed them at the end of the homework assignment as “chat topics.” Tom Barker: I’ll lead the discussion with these topics and of course keep the chat session open to any ideas, thoughts, problems you may have at any time. Tom Barker: It’s wonderful to watch the threads of thought that emerge from these few topics Tom Barker: Got the idea of how this works? Mikel Hill: Yes Vinny Tuccillo: yep Tom Barker: O, And I’ll archive the whole thing - so you can read it later Robert Parody has joined the chat. Robert Parody: hello everyone Andrew Lenzen: Somebody type something. Mikel Hill: Small talk: what’s the weather like in Roch. Tom Barker: I usually give a few minutes for the late ones to arrive Lisa Hepburn has joined the chat. Donna Perkins has joined the chat. Donna Perkins has left the chat. Mikel Hill: Well you gotta give Donna credit for trying Page 5 Leading On-Line Chat Sessions Andrew Lenzen: I’m pleased with the chat so far. I’m concerned about my connection (voice line) but it is okay speed, thus far. Laurel Harrington: Where did the “secret code” come from? Andrew Lenzen: What is the “secret code”? Tom Barker: you don’t need much speed for a chat Patrick Owens has joined the chat. Mikel Hill: Did I miss something. Joan Gregorski: Tom — while we’re waiting, what’s your take on the firestone-ford tire problem? Laurel Harrington: I see machine language or something. Tom Barker: ISO / QS does not mean quality! Laurel Harrington: Amen to that! Andrew Lenzen: Good point. What is this “secret code”? Tom Barker: That Decater Plant has a big sign outside of it proclaiming 14000 & 9002! Joan Gregorski: So they’re going to be a case study in some future CQAS quality class? Tom Barker: This course is where you will learn how to infuse quality Bryan Winkelman has joined the chat. Tom Barker: Did you get to look through my little presentation on the WEB on how DL works? Terrance Antinora: quickly Laurel Harrington: I was able to view it but the bottom lines were truncated. Joan Gregorski: Yes, I didn’t quite understand how the team stuff would work — are you going to divide us into teams?? Tom Barker: Later in the Quarter we will have a team project Mikel Hill: Did as much as humanly possible the past few days. Had a lot of good make up time shile sitting on the runway and watching a airplane engine catch on fire. Tom Barker: Mikel, what engine fire? Terrance Antinora: do we choose teams or you Tom Barker: You pick you teammates Lisa Hepburn: Excellent, I would like to work with my fellow Corning employees. Johnny Terry has joined the chat. Johnny Terry has left the chat. Joan Gregorski: works for me lisa, anybody else in the class from corning? Laurel Harrington: Is Donna still trying to join us? Tom Barker: Don’t worry about the teams right now Lisa Hepburn: A question about F tests, we were taught that to perform an F-test you must verify that the data is normally distributed. How can you compare the variance if the data is not from a normal distribution? Page 6 Leading On-Line Chat Sessions Mikel Hill: Terry and Bryan, do you want to be a team? Tom: just the plane in front of us at O’hare yesterday. No big deal if you don’t mind evacuating your flight. Terrance Antinora: no Terrance Antinora: just kidding Mikel Hill: Thanks Johnny Terry has joined the chat. Terrance Antinora: sure Tom Barker: I know you are probably swamped with reading, and viewing, but you should read the messages I post in the course conference. Thingds like the WEB FC won’t work in chats might keep disasters from happening Joan Gregorski: Oh, I just saw Johnny... I guess I can work with him :-) Bryan Winkelman: I guess so, even though you are from Kodak Johnny Terry: Hi All Vinny Tuccillo: Lisa - the Montgomery book indicates you should use the normal probability plot to check for constant variance. Lisa Hepburn: Gee, the Kodak team sounds like fun. Mikel Hill: I do my best with my handicaps Bryan Winkelman: We know Tom Barker: Teams work with people all over the world. I had a team in Turkey, Chicago, and Buffalo and they did very well! We are a global community! Timothy Jenkins has joined the chat. Terrance Antinora: i heard Mikel Hill: Come join us, however two of us work for the competition Timothy Jenkins: Hello Andrew Lenzen: I saw that a couple of you are from California. If you want to form a team, I”m game , I live in Mountain time so the difference will only be one hour if we need to chat or call on the phone. Vinny Tuccillo: Lisa - I just re-read your question. You wanted to know about normality and I was referring to constant variance - sorry. Joan Gregorski: I’ve had statisticians tell me that the test for normality is so strict that often times you can assume normality even if the test indicates it’s not exactly appropriate — Ive seen folks eyeball the plot and say it’s good enough Lisa Hepburn: Yes, we’re having too much fun to talk statistics, it’s great. Tom Barker: Let’s begin the chat. Have you all seen the first tape? Mikel Hill: How many of you actually perform DOE’s already? Bryan Winkelman: Yes Andrew Lenzen: Yes, I’ve seen it Vinny Tuccillo: Yes Joan Gregorski: yes Laurel Harrington: I haven’t seen the tape. Received it in the mail today. Page 7 Leading On-Line Chat Sessions Robert Parody: Yes Patrick Owens: Yes, I’ve seen the tape. Brian Tinch: Yes Terrance Antinora: tonight Timothy Jenkins: Yes Jeffrey Nickerson: Still waiting on materials Johnny Terry: No, it’s on back order for me Allal Ahmamad: yes Joan Gregorski: johnny — you can borrow mine if you’d like Lisa Hepburn: I’ve seen the tape and yes I design experiments although somewhat badly. Mikel Hill: I haven’t, but I went to the lecture, couldn’t find a VCR with cables connected to the TV here. Tom Barker: There was a problem with too few copies at the bookstore. This class is the biggest I have ever taught on line! Johnny Terry: Thanks Joan, will do. Lisa Hepburn: I learned very early in distance learning to get materials as early as possible. Tom Barker: Let’s give it a shot! Tom Barker: The first topic was about the measurement of heating degree days(HDD). What is a HDD? How is it defined? Vinny Tuccillo: The difference between 65 degrees and the average temperature that day. Tom Barker: right Vinny Mikel Hill: The number of days Tom Barker: nope, Mikel Mikel Hill: I didn’t finish when I saw your answer Vinny Tuccillo: The calculation of the average temp for the given day apparently is a source of debate. Tom Barker: OK Tom Barker: right - how is the average computed? Joan Gregorski: It seemed that the avg of the min and max could give you a nonrepresentative measure depending on what the distribution of temps was throughout the day.. Tom Barker: Right on Joan Vinny Tuccillo: Which is why many people suggest taking the average of , say 24 hourly measurements. Mikel Hill: That’s a lot of data though Tom Barker: Yes, did you read over the Keith Eichner article on page 13.5 in the Viewing Guide? Page 8 Leading On-Line Chat Sessions Vinny Tuccillo: Yes - but more data means more power to make decisions. Mikel Hill: agree Joan Gregorski: Why did the author make the last statement he did — “it probably wouldn’t make a big difference of estimating heating needs over an entire season”? Terrance Antinora: and its easy to collect Lisa Hepburn: If you used the average instead of a single value would you be able to utilize the variability of the temperature throughout the day. Robert Parody has left the chat. Tom Barker: I suspect, Joan, Thet he didn’t want to upset all those years of history with a new better way of doing it! Tom Barker: Right Lisa Tom Barker: Have you encountered situations where there is a better way to measure a characteristic, but change does not take place? Bryan Winkelman: Yes, everyday at work. Vinny Tuccillo: Does a bear..... Robert Parody has joined the chat. Joan Gregorski: Would those averages of min and max follow a normal distribution over a long enough time period just like any other average?? Tom Barker: Right! “We can’t afford to upset the data base.” Lisa Hepburn: It is extremely difficult to introduce the idea of change to processes that are already implemented. Tom Barker: The min - max are only extreme samples and probably would not be normal Johnny Terry: With enough data, it can appear to be normal Mikel Hill: We measure temperature with spent air sensors, but a new method using IR pyrometers is being proposed, however the crossover may not correlate Tom Barker: Johnny, not necessarily - a deck of playing cards is always uniform Tom Barker: The distribution depends on the characteristics of the function in question. Johnny Terry: oops, you’re right Vinny Tuccillo: Now if you take the monthly average of those daily averages, and plot them, the CLT states this distribution will be more normal than the original distribution. Tom Barker: Now, you might be thinking of taking samples and averaging the observations in those samples - then the distribution of the sdample averages will tend to normality (CLT consequrnce #1) Tom Barker: But back to the HDD Joan Gregorski: that’s just an x-double bar term Tom Barker: I wanted to make this point that goes beyond the use of the HDD in the teaching example to get you thinking about response variables. Bringing the chat back on track Page 9 Leading On-Line Chat Sessions Tom Barker: …and representative sampling. Alan Janes has joined the chat. Tom Barker: If the response is messed up, the experiment might also be messed up! Mikel Hill: I’ve seen that! Tom Barker: Do you have any more insights? ON HDD? Tom Barker: If not, then to the next topic! Tom Barker: If I have two normal distributions, and take the difference, what will the shape of the resulting distribution be? Johnny Terry: Tom, if your reponse variable is messed up, what should you do Terrance Antinora: normal Joan Gregorski: normal Lisa Hepburn: A normal distribution Vinny Tuccillo: normal Tom Barker: Fix the response first! Mikel Hill: normal Bryan Winkelman: Normal Tom Barker: Right on all counts Terrance Antinora: check for interaxtions Terrance Antinora: c Tom Barker: well, Terrance, interactions are an influence on the response - so if the response is bad, the interactions are probably not helpful Vinny Tuccillo: Wouldn’t the difference between any 2 distributions - normal or not be normally distributed? Robert Parody has left the chat. Tom Barker: Yes, Vinny it is normal - we’ve been having a little of a hangover from the last topic on response variables. Vinny Tuccillo: Oh - now that I’m back in college, I should have known there would be hangovers... Robert Parody has joined the chat. Tom Barker: I expected that response from sombody! Tom Barker: OK back to the difference between two normals Lisa Hepburn: The difference between, say a, a lognormal distribution and a normal distribution would be a normal distribution. Lisa Hepburn: That last one was meant to be a question. Tom Barker: I’m not sure about LogN and N Tom Barker: Lisa where do you encounter log normals? Johnny Terry: How about a chi square and normal (difference) Lisa Hepburn: It is a good fit for splice loss measurements where a value of less than 0 is not possible. Johnny Terry has left the chat. Page 10 Leading On-Line Chat Sessions Tom Barker: are you into pasrticle size, Lisa? Tom Barker: particle Lisa Hepburn: Yes, I’m familiar with particle size. Tom Barker: OK, Let’s not get esoteric on didderences between different distributions Tom Barker: differences Joan Gregorski: when you do a t-test, you’re basically using a zero difference distribution between 2 normals as your null distribution Tom Barker: right, Joan - tghat’s why I posed the chat toipc Tom Barker: I want to take that formula apart, so you really understand what it’s doing. Joan Gregorski: if your test statistic falls within the distibution based on the cutoffs determined by alpha, you assume there is no difference between the means, otherwise you assume there is and take on that risk? Tom Barker: right Vinny Tuccillo: the risk associated with the p-value? Tom Barker: careful of p values Vinny Tuccillo: That’s why I ask. Vinny Tuccillo: Or are you referring more to the alpha and beta risk? Tom Barker: I wrote to Robert Parody in the Q&A conference on the problems with p values. Joan Gregorski: p value is the probability you’ll get your test statistic value if the null hypothesis is true Vinny Tuccillo: My interpretation of p-values has always been similar to what Joan just typed. Robert Parody has left the chat. Tom Barker: Have you read my thing on the problems with p values? Robert Parody has joined the chat. Johnny Terry has joined the chat. Vinny Tuccillo: Poor Johnny. Tom Barker: Have you read my thing on the problems with p values? Vinny Tuccillo: not yet Andrew Lenzen: Where is it? Tom Barker: They are statistics - subject ot variation. They are not values you can depend on. Tom Barker: I wrote to Robert Parody in the Q&A conference on the problems with p values. Vinny Tuccillo: Just pulling it up now. Joan Gregorski: You don’t obtain a p-value until you complete the t-test; it’s not a criteria involved in performing the test but a result.. Page 11 Leading On-Line Chat Sessions Tom Barker: It was his question on hw #1 The Q&A is a good source for answers to questions you might want to ask and might not think to ask Laurel Harrington: tom-you mentioned taking a formula apart a minute ago. Which formula are you referring to? Tom Barker: The t test for the difference between two means Lisa Hepburn: Yes, I read your discussion of p values. It was enlightening since they were drilled pretty hard in the Fundamentals course. Tom Barker: Right Joan Robert Parody: Are you saying that the p-value given by the computer may change for the same t-stat? Johnny Terry: the same applies to t-test Tom Barker: For the same data, the p value will be calculated correctly by the computer - but if you did the expewriment over again and had different data, the p would be different Joan Gregorski: right, because your test statistic would differ every time based on your data.. Tom Barker: I suggest that you read my viedws on p values in the Q&A conference. Tom Barker: Right Joan Tom Barker: Let’s get back to the t test formula for the difference between two means Robert Parody has left the chat. Robert Parody has joined the chat. Tom Barker: In the t test formula, we pool the variances. Are Standard deviations additive? Terrance Antinora: no Vinny Tuccillo: No - variances are additive Timothy Jenkins: No Brian Tinch: No Johnny Terry: no Bryan Winkelman: no Lisa Hepburn: I have to agree Robert Parody: Tom, I missed you response to my question about the p-values. I am having a problem saying connected. Andrew Lenzen: variances are Tom Barker: Robert, I told the class to read my reply to you on this matter Tom Barker: right std. dev.’s are not additive Tom Barker: OK, how about variances - are they additive? Vinny Tuccillo: yes Jeffrey Nickerson: yes Mikel Hill: Yes Page 12 Leading On-Line Chat Sessions Brian Tinch: Yes Jodie Watson: yes Lisa Hepburn: YUP Laurel Harrington: yes Allal Ahmamad: yes Tom Barker: You should distinguish between population variances and sample variances. Bryan Winkelman: Yes Joan Gregorski: yes Terrance Antinora: yes Tom Barker: You should distinguish between population variances and sample variances. Tom Barker: Right - almost. Population variances are additive. But sample variances can come from different sample sizes. Would you give as much weight to a variance from 5 observations as a variance from 30 observations? Lisa Hepburn: No Laurel Harrington: no Andrew Lenzen: Good point. No, I would not Jeffrey Nickerson: no Vinny Tuccillo: Good point Patrick Owens: No Bryan Winkelman: No Terrance Antinora: true Joan Gregorski: no Robert Parody has joined the chat. Allal Ahmamad: no Tom Barker: So how do we handle it? Johnny Terry: would you pooled them Terrance Antinora: pool Mikel Hill: 30 is a minimum number typically Bryan Winkelman: Pool them Robert Parody has left the chat. Joan Gregorski: correct for it via CLT Robert Parody has left the chat. Tom Barker: No, Joan, the CLT does not come into play here Tom Barker: “pooling” is an expression for doing a weighted average of the sample variances. Tom Barker: We would weight the 30 observation set more than the 5 observation set. This is the essense of the pooling formula found on page 29 (SA33) of the viewing guide. Page 13 Leading On-Line Chat Sessions Tom Barker: You should have your viewing guides out for the chats - I often refer to them Tom Barker: I’ll give you a minute to get them and look this up Robert Parody has joined the chat. Alan Janes has left the chat. Joan Gregorski: you subtract 2 in the denominator because you’ve calculated the sample? variance for both samples? Alan Janes has joined the chat. Jeffrey Nickerson: Is this it? Sp^2=((n1-1)s2^2+(n2-1)s2^2)/ (n1+n2-2) Lisa Hepburn: I have it, I though subtracting 2 was due to the df? Mikel Hill: I think you are right Johnny Terry: corrrect Lisa Vinny Tuccillo: Tom - your original question was around adding variances. However, isn’t pooling the variance more like “averaging” the variances. Johnny Terry: one df for each sample n Joan Gregorski: you have to subtract 1 df for each piece of information you’ve calculated using the data (i.e. the two sample variance terms in the numerator)?? Tom Barker: Yes, it is, but it’s doing a weighted average Tom Barker: recall the formula for the variance. p16 ise24 Vinny Tuccillo: So you would pool if you were trying to get the average variance of a number (for example 4) distributions as opposed to adding the 4 variances and dividing by 4.... Vinny Tuccillo: Unless the 4 distributions have the same sample size. Tom Barker: right Vinny Vinny Tuccillo: In which case the average of the variances would be okay. Tom Barker: exactly Tom Barker: recall the formula for the variance. p16 ise24 Laurel Harrington: I’m not sure I understand the difference between what Joan offered about the source of the “2” and the Laurel Harrington: answer Lisa gave. Joan Gregorski: i think they’re one in the same... Tom Barker: we divide by n-1 (to get an unbiased estimate of the sigma squared), so we multiply by n-1 when we turn the sample variance back into the ....? what is is called? Robert Parody has left the chat. Vinny Tuccillo: sum of squares Tom Barker: joan is right they are one in the same Tom Barker: Right Vinny Tom Barker: the sum of squares is always additive! Page 14 Leading On-Line Chat Sessions Robert Parody has joined the chat. Tom Barker: Keep this in mind for next week when we get into ANOVA Tom Barker: Robert, have you reied setting your buffer in FC to 64000? That might keep the disconnects from happening Tom Barker: tried Tom Barker: Give the DL techie crew a call tomorror and see if they can help you out with your connection problem Tom Barker: Back to statistics Tom Barker: So in pooling, the thing we are adding is the sum of squares, and sum of squares are ALWAYS additive! Tom Barker: You can think of it either way. Sum of squares are always additive, or we are adding the variances in a weighted manner. What ever works for you. Tom Barker: If we go further with the t test formula, we see that the pooled std. dev. is then multiplied by the square root of the sums of one over the sample sizes. Did you ever wonder where that came from? Tom Barker: This is in SA34 Timothy Jenkins: That is the same thing we do to calculate std error of the mean. Tom Barker: yes Vinny Tuccillo: CLT? Daniel Blodgett has joined the chat. Tom Barker: This is the CLT in action Tom Barker: See SA34 P 29 Robert Parody has left the chat. Tom Barker: Want to know more? Laurel Harrington: yes Tom Barker: Good, I want to tell! Vinny Tuccillo: If you’re offering, what will our stock price be at close tomorrow? Jodie Watson: CTL=Central Limit Theorm Robert Parody has joined the chat. Tom Barker: Yes, Jodie Tom Barker: Stock will be up 7/16ths Vinny Tuccillo: thanks for the tip - want to put a CI around that? Mikel Hill: .001 Tom Barker: Let’s start with a diversionary example that will lead to the answer and also check on another idea that is important. By the way, I have an animated presentation in the session #1 conference that goes through this. Tom Barker: If I have two buckets of parts that will be joined to each other in an assembly, and I’ll take one part from bucke A and one part from bucket B, what will the average assembly length be if… Tom Barker: Bucket A Bucket B A ‘mini-lecture’ SA34 is a slide in the viewing guide I have this ready to copy and paste Page 15 Leading On-Line Chat Sessions Tom Barker: x bar 2" 3" Tom Barker: s 0.2 0.1 Timothy Jenkins: 5" Joan Gregorski: avg will be 5 Jeffrey Nickerson: 5" Mikel Hill: 5" Terrance Antinora: i saw this 5" Lisa Hepburn: 5" Bryan Winkelman: 5" Tom Barker: Right Tom Barker: Ok, now the harder part. What will the standard deviation of the assemblies be? Jeffrey Nickerson: .2236 Mikel Hill: .15" Tom Barker: Jeffrey Right - how did you do that? Robert Parody has left the chat. Tom Barker: Mikel - not right Jeffrey Nickerson: square root of the some of squares Bryan Winkelman: Your off our team Mikel Bryan Winkelman: Just kidding. Tom Barker: What is that approach called Jeffrey? Jeffrey Nickerson: ????? Jodie Watson: SQRT the pooled variance Tom Barker: Additive tolerance stack up Tom Barker: The formula says to add the variances (we assume equal sample sizes) and take the square root of this sum. SO: s overall = sqrt(s^2 sub A + s^2 bub B) Tom Barker: “^” means raise to a power Tom Barker: Many engineers would simply add the two std.’s Tom Barker: So what does this have to do with t tests? Tom Barker: Back to SA34 Mikel Hill: As I did, and I worked in this area for a while. That’s why I changed to research Tom Barker: In the t test for the difference between two means, we are finding the “linear combination” of two populations. When you do so, we need to add the variations - like in the additive tolerance example. Tom Barker: Variances must always be added (never subtracted) even if we are finding the difference between the two means. Why do we add variances? (clue can variances be negative?) Mikel Hill: that would create an imaginary number Tom Barker: Why do we add variances? Page 16 Leading On-Line Chat Sessions Daniel Blodgett: Variance goes in both directions from the mean Terrance Antinora: sqared number can’t be neg. Vinny Tuccillo: variances cannot be negative Patrick Owens: No negative variances. Laurel Harrington: yes they can be negative but squares make positive numbers. Tom Barker: Right squared numbers are not allowed to be negative! Tom Barker: It’s nice to have mathematics help us at times! Tom Barker: So, we have the two variances (now represented by the pooled common estimate) and they are added –– but we are looking at the difference between two averages. Tom Barker: The central Limit theorem comes in with its second consequence and we have the standard error of the mean Joan Gregorski: so we’re using the concept of additive tolerance and correcting for samples sizes? Tom Barker: No, not correcting Tom Barker: Utilizing thr CLT Tom Barker: So the formula has the sum of the variances each divided by the number of observations in each average and then we take the square root to get the std. dev. The common pooled variance is taken out of the radica Tom Barker: radical Tom Barker: I suggest that you download the presentation on this concept - so you know what the formula for the teo sample t test is all about. Tom Barker: two Johnny Terry: Bye all, I have to catch my flight out (Atlanta) Tom Barker: The difference between this course and maybe a short course you have taken on DOE is, we get into the concepts behind the formulas, not just in the proper use of the formulas. Tom Barker: Have a good flight - we onlt have one more topic left Tom Barker: only Johnny Terry: I have a bout 5 minutes Tom Barker: I want you to understand the ideas as well as being able to use the material properly. Tom Barker: One last chat topic Look at the t Table on page 35 in the viewing guide. Andrew Lenzen: ok Tom Barker: For a 5% two tail risk (middle of the Table) what is the t value for 1 df? Bryan Winkelman: 12.706 Joan Gregorski: 12.706 Terrance Antinora: 12.706 Brian Tinch: 12.706 Page 17 Leading On-Line Chat Sessions Timothy Jenkins: 12.706 Lisa Hepburn: 12.705 Patrick Owens: 12.706 Laurel Harrington: 12.706 Andrew Lenzen: ditto Tom Barker: good! Tom Barker: Now go to 2 df Daniel Blodgett: 12.706 Laurel Harrington: 4.3027 Terrance Antinora: 4.3027 Timothy Jenkins: 4.3027 Andrew Lenzen: 4.3027 Vinny Tuccillo: 4.3 Brian Tinch: 4.3027 Bryan Winkelman: 4.3027 Jeffrey Nickerson: 4.303 Lisa Hepburn: 4.3027 Tom Barker: Quite a change for one more observation! Lisa Hepburn: Yes Tom Barker: Good return on investment\ Patrick Owens: 4.3027 Tom Barker: How about for 4 df? Lisa Hepburn: 2.7764 Bryan Winkelman: 2.7764 Laurel Harrington: 2.7764 Terrance Antinora: 2.7764 Timothy Jenkins: 2.7764 Brian Tinch: 2.7764 Jeffrey Nickerson: 2.776 Vinny Tuccillo: 2.78 Andrew Lenzen: 2.7764 Patrick Owens: 2.7764 Tom Barker: YEP! Tom Barker: and for 5df? Lisa Hepburn: 2.5706 Jeffrey Nickerson: 2.571 Bryan Winkelman: 2.5706 Andrew Lenzen: 2.5706 Laurel Harrington: 2.5706 Daniel Blodgett: 4.3027 Page 18 Leading On-Line Chat Sessions Timothy Jenkins: 2.5706 Vinny Tuccillo: 2.57 Mikel Hill: 2.571 Brian Tinch: 2.5706 Tom Barker: We’re getting into diminishing returns Joan Gregorski: as your df increase, the ability to show a significant difference gets easier.......... Tom Barker: Right! Tom Barker: So, if you had to come up with a preliminary number of observations to calculate a std. dev. to determine the final sample size, where could you start? How many observations minimum? Laurel Harrington: 3 Patrick Owens: 2.5706 Terrance Antinora: probably about 2 or 3 Bryan Winkelman: 3 Tom Barker: Right Laurel! Tom Barker: 2 is not enough - only 1df Tom Barker: where diminishing returns were taking place. around 5 observations. Lisa Hepburn: Gee, I just did an experiment and did 20 samples to get a feel for the standard deviation. Tom Barker: probably too many Lisa Hepburn: I see, what a waste of my time! Daniel Blodgett: 2.5706 Terrance Antinora: time is of the essence Tom Barker: Remember EFFICIENCY is what design of experiments is all about Laurel Harrington: Why is 30 used as a sample size with such frequency? Daniel Blodgett: 3-5 Vinny Tuccillo: Laurel - CLT Johnny Terry: Bye all Joan Gregorski: because you can assume a normal distribution if you have a sample size above 30?? Johnny Terry has left the chat. Tom Barker: Laurel, that’s because if you want to be able to see a pattern in a distribution, you need 30 or so observations to get enough “bulk” in the sample Vinny Tuccillo: when n=30, your subgroup averages arealmost sure to be normally distributed Tom Barker: No Joan Terrance Antinora: easier to see a distribution Tom Barker: I think this may be where you get that idea, Joan Page 19 Leading On-Line Chat Sessions Tom Barker: look at the bottom of the t Table Jeffrey Nickerson: Most T-tables end at df 29 Tom Barker: for 30 df t = (same 5% two tail risk) Daniel Blodgett: 2.0423 Tom Barker: What’s the t? Vinny Tuccillo: 2.0423 Tom Barker: right Patrick Owens: 2.0423 Tom Barker: What about at infinity df? Daniel Blodgett: 1.96 Terrance Antinora: 1.96 Brian Tinch: 1.96 Vinny Tuccillo: As n goes to infinity, t approaches Z Lisa Hepburn: Isn’t that the same as a normal distribution. Patrick Owens: 1.960 Allal Ahmamad: 1.96 Timothy Jenkins: 1.96-same as a z test Tom Barker: By the way that’s from the “Normal” distribution - remember Gossett? Tom Barker: The difference between 30 df and infinite df is about only 4% Tom Barker: So we can see that the t has converged to the normal at about 30df. Tom Barker: This does NOT mean that you get a normal distribution at 30 observations. It only means that the critical values of the t and normal are about the same for 30 df. See ISE04 (on the video where it moves). - Joan this is where you Tom Barker: were duped! Joan Gregorski: Yep, and I’ve proved it to myself at work where I have large data sets and the normal assumption is DEFINITELY NOT a good one. Lisa Hepburn: So your sample distribution should still be verified for normallity before you run a t-test if you have 30 or more samples. Tom Barker: For the t test, you are comparing means and means ARE distributed normally (CLT #!) so you don’t need to worry! Tom Barker: I think I’ve got you all thinking - good that”s my job! Tom Barker: BUT I don’t know about you, but I’m tired that’s it for tonight - unless you have questions. Vinny Tuccillo: Thanks Lisa Hepburn: I’m OK, thanks and have a good night. Tom Barker: Except ONE LAST THING Joan Gregorski: I have a question: I have datasets where each sample has a very large number of measurements associated with it. The df’s then are so large that I always come up with a significant difference between samples but I think it’s jus Laurel Harrington: I need to digest some of this first... Page 20 Leading On-Line Chat Sessions Joan Gregorski: a result of the number of observations.. Brian Tinch: Thank You Tom Barker: we have some students on the west coast who can’t getr to the chat at 7 Tom Barker: could we put the chat off until 7:30 each week to accomodate them? Laurel Harrington: ok for me Timothy Jenkins: Yes, that would help Mikel Hill: I’m ok with that. Lisa Hepburn: OK for me. Brian Tinch: 7:30 works better for me Bryan Winkelman: No problem for me Jodie Watson: oK! Vinny Tuccillo: Works for me - I’m on West coast Allal Ahmamad: ok Terrance Antinora: sure Joan Gregorski: OK with me, but I get pretty cranky about 9pm :-) Andrew Lenzen: Works for me. Tom Barker: we’ll stop at 9pm Patrick Owens: Yes. Lisa Hepburn: Your pregnant Joan, you’re bound to be a little cranky. Tom Barker: or Joan can exit Tom Barker: . Tom Barker: . Mikel Hill: . Bryan Winkelman: . Terrance Antinora: . Lisa Hepburn has left the chat. Tom Barker: Then I will reschedule the chat on Tuesday Lisa Hepburn has joined the chat. Jodie Watson has left the chat. Lisa Hepburn has left the chat. Mikel Hill has left the chat. Allal Ahmamad has left the chat. Jeffrey Nickerson has left the chat. Bryan Winkelman has left the chat. Vinny Tuccillo: Are we done? Terrance Antinora has left the chat. Brian Tinch has left the chat. Laurel Harrington has left the chat. Page 21 Leading On-Line Chat Sessions Space For Your Notes What did you see happening in this chat? LDTATNC IC01 Insight Infrastructure Tool Space For Your Notes Could you see the Learning? LDTATNC IC02 Insight Infrastructure Tool What you have learned: Space For Your Notes You have seen some of the alternative ways to involve Distance Learning students in discussions and create a more active learning environment. A sampling of comments by the customer (students) shows there is a disdain for threaded discussions, while other comments on synchronous chats indicate a value added effect. Resources requirements for the synchronous chat also seem to be less intensive for the faculty and while a bit more for students who are prone to “do the least for the grade” (those students who are in school merely for grades, not learning), the learning to resource ratio is most likely much higher. LC01 Insight Infrastructure Tool Page 22 Leading On-Line Chat Sessions Space For Your Notes Logistics of Leading a Chat LC01 Insight Infrastructure Tool Space For Your Notes Have the topics and the archived chat from before open LC02 Insight Infrastructure Tool Space For Your Notes Chat Room Topics LC03 Insight Infrastructure Tool Page 23 Leading On-Line Chat Sessions Space For Your Notes Copy from Archive LC04 Insight Infrastructure Tool Space For Your Notes Saves Lot's of keystrokes! LC05 Insight Infrastructure Tool By preserving your “Brilliant” interjections from the past, you are able to keep the chat on track and inject learning easily. LC06 Insight Infrastructure Tool Space For Your Notes Page 24 Leading On-Line Chat Sessions Space For Your Notes Of course let the chat digress and create new responses in the process LC07 Insight Infrastructure Tool Space For Your Notes What do You Think? • Too rigid? • I think I'll try it! • Need more spontaneity LC08 Insight Infrastructure Tool Archiving A Chat Space For Your Notes • • • • • Do not close chat room Wait until all have left ‘Select All’ in chat session area Copy Create new message in session conference • Paste into new message LC09 Insight Infrastructure Tool Page 25 Leading On-Line Chat Sessions Space For Your Notes • Select all in chat session area • Copy LC10 Insight Infrastructure Tool Space For Your Notes Create new message in session conference Insight Infrastructure LC11 Tool Space For Your Notes Paste into new message LC12 Insight Infrastructure Tool Page 26 Leading On-Line Chat Sessions Space For Your Notes Send Message LC13 Insight Infrastructure Tool Space For Your Notes Move message into chat archive. LC14 Insight Infrastructure Tool What you have learned: ou have learned about the logistics of a chat in FirstClass. YouFor canYour expedite Space Notesthe entry of your instructor inputs by having them available in previously recorded messages or from chat archives from other years. Besides making fewer keystrokes for your entry, this assures that consistency in learning takes place between offerings of the course. You may certainly improvise as the need arises and in doing so create even more “Brilliant” inputs for this class and for future classes. Y You have seen how an archive is created by copying the contents of the chat window and pasting them into a message that is then posted to the chat archive for that session. In this way, the students who were at the chat can review the learning (using the search option of FirstClass). AND students who could not attend are given the opportunity to review what they missed. (..and you can check if they did, by looking at the message FCI23 FCI23 ‘History’). Insight Infrastructure Tool Page 27