# Messages

Don't be worried or alaramed, regarding the "bring a simple calculator for your exam" part of your practicalities & preparations. You might need to compute \sqrt{49} og \log(1 + 3.33) or 5.55/4.44, but nothing superdrastic. There's a webpage with semiprecise information regarding which calculators are *not* allowed for use for the exams, so follow these instructions:

ttps://www.uio.no/studier/eksamen/hjelpemiddel/mn-math-kalkulatorer.html

As announced earlier, each candidate can bring to her or his exam table *one page of handwritten notes, in the student's own handwriting*.

I'm aware that I in one of my previous messages have written "one sheet of paper* (as below), but this is *not* to be interpreted as "two pages of the same sheet of paper*. Again: one page only.

"One more thing, regarding the four-hour exam: Students will be allowed (by me) to bring with them (a) a simple calculator, with no memory, just enough to compute \sqrt{49} and a bit of \log and \exp, perhaps; (b) *one sheet of paper, in their own handwriting* (where they may write anything they want, from E = mc^2 to a formula for the mean of a normal distribution with mean zero)."

1. On Mon Nov 19 I went through the Oblig, with time also used for the martingales, etc., and the exercise with n = 238 drug users in clinics, via Ch 10 hitting-time models; cf. R script com31b. I also started "repetisjon", looking back at our curriculum.

2. For Mon Nov 26, do these two exercises. (a) The drug users, once more, with Aalen-Gjessing model, of the type X_i(t) = c_0 \exp(x_i \beta) - (\mu_0 + \gamma x_i) t + W_i(t) reaching zero. (b) Suppose an event takes place in one of ten possible ways, corresponding to independent times T_1, ..., T_{10}, with hazard rates \alpha_1(s), ..., \alpha_{10}(s), and that the event happens when the first of these sub-events takes place. Find the hazard rate and survival curve for the time T to event.

3. On Mon Nov 26, we're in repetisjonsmodus, looking through our curriculum. Prepare, by having questions, concerning details or the bigger picture.

Thanks for your valiant efforts with the Oblig! On Mon Nov 19 I will go trough these two exercises. Also, do Exam 2016 #3, plus one more exercise -- which I haven't had time to attempt myself yet, but I hope it's doable. Visit our friends the n = 238 Australian drug users, once again, and make a Chapter 10 type model for the length of time an individual spends in a clinic, factoring in the covariate x = methadone use. Your model should take the time t_i to be the crossing-time for level zero for a process of the form X_i(t) = c_i - \mu t + W_i(t), with Brownian motions W_i, with the same \mu for all, but with c_i = c_0\exp(\beta x_i). Program the log-likelihood function, find ML estimators for c_0, \beta, \mu.

I will otherwise round of the curriculum parts of Ch 10.

1. On Mon Nov 12 I'm partly busy with the FocuStat Research Kitchen (the "Kombo Kjøkken", on how to combine information from diverse sources). So on this particular Monday I will teach *from 12:30 to 13:30* (but we're doing fine overall, regarding time).

2. Exercises: Exam 2016, Problems 1 and 2. Also: go to old Egypt, and try to fit the time-to-hit-zero models with X(t) = c - mu t + W(t), with W(t) Brownian motion.

3. And of course, deliver your Oblig reports in time, Tue Nov 13 by 13:59, preferably much earlier!

1. On Mon Oct 29 I used time to go through all of Exam 2012, in particular the occurrence / exposures things for the four Danish cities and five age groups. I also more or less rounded off Ch 6.

2. The Oblig is out; deadline for handing in your reports, as a pdf to the devilry system, is Tue Nov 13 at 13:59. Enjoy.

3. On Mon Nov 6, do Exercises 2 and 3 from the Exam 2014 set. The first of these adds some insights to Ch 6 material.

4. Riccardo Parviero gives us a story on the survival of Fake News stories.

5. I start, gently, with the last chapter from the curriculum list, namely Ch 10.

The pensumliste for the stk 4080 course, autumn 2018, is as follows. First, from the Aalen, Borgan, Gjessing book:

Ch 1: read all of it (though it's not "technical", so to speak).

Ch 2: 2.2 (and check related Nils exercises).

Ch 3: 3.1, 3.2, 3.3, with a more cursory look through 3.4.

Ch 4: 4.1, 4.2.

Ch 5: all of it (and check related Nils exercises).

Ch 6: 6.1, 6.2, 6.3, 6.4.

Ch 10: 10.1, 10.3.

In addition, all exercises we've been through during the course, from the "Nils Collection" and from the book, are defined as active curriculum.

1. On Mon Oct 22 we listened to PhD student Jonas Schenkel's story on Cox with lasso. I did Exam 2010 Exercises 2, 3, and started Ch 6.

2. A pensumliste, curriculum list, will be put up here reasonably soon: parts of Chapters 1, 2, 3, 4, 5, 6, 10, plus all Nils Exercises we've been through.

3. For Mon Oct 29, try to do all of Exam 2012, i.e. Exercises 1, 2, 3. The first is essentially material from the book, and the second is partly from the start of Ch 6. The more demanding one is the third. Try absolutely to set up likelihoods, etc., where the 2012 text says "you don't need to do this now". Also go through the numerical example there, with occurrence / exposure lung cancer data for four Danish cities.

1. On Mon Oct 15 I discussed material for Aalen's linear hazard rate regression model (Ch 4), and also parametric regression models (Ch 5). I went through exercises, including the famous Australian drug users, also via parametric models of the type \alpha(s,\theta)\exp(x_i^\tr \beta).

2. PhD student Jonas Schenkel wil use 25 minutes on Mon Oct 15, to tell us about lasso for Cox.

3. An extra half hour, for those who want to, Mon Oct 22 (I have piano lessons every second Monday, so on the other Mondays I can offer such extra half-hours).

4. I'm uploading my com27b script, which does a somewhat simple version of the Aalen linear hazard regression model.

5. For Mon Oct 22, first do the Australians again, with x = methadone dose as the single covariate, but for clinic I and clinic II separately. Compute and display the B_{1,clinic1} and B_{1,clinic2} curves. Compare them and test if they are equal.

Then go to *earlier exam sets*, now posted on the co...

On Mon Oct 22 we take another extra half-hour-or-so, from 15:15, in the UE 32 room -- for those who want to. See previous similar message for Mon Oct 8.

An earlier message, given a few weeks ago, read as follows:

"The Oblig (obligatory exercises) dates are now set. On Thu Nov 1 the set of exercises is made available; by Mon Nov 12, before 12:59, you hand in your reports. This needs to be done via the Devilry System [sic]; more details regarding this later. Expect both theory and practice, so to speak."

Apparently certain rules imply that the Oblig should have *two weeks of time*, which means that I'm now revising the above to (i) starting date Tue Oct 30, (ii) deadline for handing in report Tue Nov 13. More details will come later.

One more thing, regarding the four-hour exam: Students will be allowed (by me) to bring with them (a) a simple calculator, with no memory, just enough to compute \sqrt{49} and a bit of \log and \exp, perhaps; (b) *one sheet of paper, in their own handwriting* (where they may write anything they want, from E = mc^2 to a formula for th...

1. On Mon Oct 8 I went through more on Cox regression and also Aalen's linear hazard rate regression model. For those having wished for more time and more details we had 45 extra minutes in UE 32, with a bit on martingales, etc.

2. I've uploaded com18a (on the pornoscope data) and com21c (on Cox regression for the Australian drug users) to the website. The latter also has my simple code for the log-partial-likelihood, for this dataset, finding the same beta estimates and standard errors as the "coxph" command in R's "survival" package.

3. Next week I do more with the Aalen linear model, and round off Ch 4. I also start Ch 5, which should contain non-surprising material, partly in the category of "similar, but simpler, than Ch 4".

4. For the n = 238 Australian drug users, do *both* semiparametric Cox regression *and* two parametric versions, with (i) the Weibull \theta s^\gamma and (ii) the Gomertz \theta \exp(\gamma s) for the baseline hazard. These two...

The two student representatives and I have had a pleasant meeting, and we agreed it could be useful with *occasional extra time*, with more room for discussion of details, of exercises, etc. So *once in a while* we'll have an extra half hour, hopefully in the same room VB 123, from 15:15 to around 15:45, *for those wishing to take part*. I promise not to go through new wonderful theorems during such extra half hours, and similarly I will not go through separate exercises then -- so this is not at all mandatory, and everyone is free to leave at the verkerkian time 15:03. It's rather a simple offer, a tilbud, to those wishing more time for questions-and-answers, more detailed explanations of some of the material we go through, etc.

1. On Mon October 1 I went through the basics of the Cox regression model, and used time to illustrate concepts, results, interpretation, with the dataset on 238 Australian drug users. Then we ran away to listen to Ingrid Van Keilegom's talk ... on survival analysis.

2. Next week we will discuss the second most important model and method of Ch 4, namely Aalen's linear hazard rate regression model.

3. I have a couple of R scripts to upload, regarding pornscope data and the Australian drug users, and shall do this pretty soon.

4. Our two famous PhD students, Jonas S. and Riccardo P., will give 25 minute lectures, on Mon Oct 22 and Mon Oct 29; they will decide who lectures when. Their lectures will be defined as curriculum for all (and pdfs will be put on the course website).

5. Importantly, there's now a new version of the Nils Exercises and Lecture Notes (version 0.59, as of 2-x-2018, now 42 pages); you...

1. On Mon Sep 24 I went through the essence of Nils Exercises 24, 25, 26, 27, including analysis of the pornoscope data. I then rounded off the material of Ch 3, and started on Ch 4. The datasets for both of Chs 4 and 5 are of the form (t_i, \delta_i, x_i), with time-to-event (possibly censored), an indicator for non-censoring, and covariates. On Mon Oct 1 I go further in Ch 3.

2. For Mon Oct 1, carry out the following exercise (which suitable edited will land in the Nils Collection). Go to the "some datasets" part of this website, and upload the dataset "heroin2-data" to your computer. It contains six columns, (id, tt, delta, x1, x2, x3), with explanation given in the "some datasets" part, for n = 238 Australian drug users. You should now carry out basic Cox regression for tt, the length of time a user stays in clinic, with respect to x1, x2, x3. Do this via

library("survival") # this is a package

well <- coxph(Surv(tt,delta) ~ x1 + x2 + x3)

summa...

If queries or thoughts or views about the course or how it's being taugt etc., please contact me directly (!) -- but it's also perfectly ok to go via one or both of the two student representatives we're now blessed with:

** Eva Steine Dahl / evasda@uio.no (or evasda@student.matnat.uio.no, I presume)

** Vegard Stikbakke / vegarsti@student.matnat.uio.no

1. On Mon Sept 17 we were pressed for time because of the seminars, but I went through material of the 2nd half of Ch 3, along with the essence of Exercises 4, 7, 8, 12 from Ch 3.

2. I've extended the Nils Exercises and Lecture Notes to version 0.44, currently 33 pages. Print out a set for yourself.

3. The Oblig (obligatory exercises) dates are now set. On Thu Nov 1 the set of exercises is made available; by Mon Nov 12, before 12:59, you hand in your reports. This needs to be done via the Devilry System [sic]; more details regarding this later. Expect both theory and practice, so to speak.

4. On Mon Sept 24 I use half an hour to round off Ch 3 properly, with a bit more om testing, on confidence bands, on event histories involving more than two boxes.

5. Exercises for Sept 24: Nils Exercises 24 (Wald ratios), 25 (inference for quantiles), 26 (simultaneous confidence bands), 27 (pornographic data).

1. On Mon Sept 10 I went through various exercises, from ABG Ch 3, and the main material for Ch 3, including Nelson-Aalen, Kaplan-Meier, martingales, a bit om tests and confidence bands, and approximate normality.

2. There's an updated 23-page version 0.33 of the Nils Collection of Exercises and Lecture Notes; print it out (and read through and think about also those exercises we don't go through in class).

3. I'll upload an R script for the analyses related to Exercises 3.1-3.2.

4. On Mon Sept 17 there will only be two hours of teaching -- since I'm called for at a statistics seminar with Idris Eckley (usually these seminars are on Tuesdays, but on this particular occasion it's on a Monday).

5. Exercises, from Ch. 3: 4, 7, 8, 12. Also, follow up Exercises 1 and 2 by constructing, evaluating and interpreting a real test for the hypothesis that the new drug doesn't work better than placebo.

1. On Mon Sept 3 I started Ch 3, on the famous nonparametric Nelson-Aalen and Kaplan-Meier, and also about the martingales, variance processes, etc. We also did Nils Exercise 8 on IUD data, and more.

2. I've uploaded com14c to the website, with the IUD expulsion analyses, including how to ML estimates for the two-parameter model.

3. I'm in the process of writing up Nils Exercices 9, 10, 11, 12, 13, 14, and the updated document will be uploaded to the course site within a few days. It will conveniently sum up the salient features of martinales, limiting normality, etc.

4. Exercises for Mon Sept 10: First, do Nelson-Aalen, with bands, for the men and women of Ancient Egypt. Plot the difference curve D(t) = Awomen(t) - Amen(t), with a confidence band, and summarise what you learn. Then do Exercises 1, 2, 3 from ABG Ch. 3. Finally, do a Nelson-Aalen confidence band for the IUD expulsion data, though it becomes rath...

I've created the group "STK 4080-9080 Autumn 2018: Survival Analysis" on Facebook: https://www.facebook.com/groups/236121410385506/

This group is a forum for students and other followers of the UiO course STK 4080-9080 Autumn 2018. It's meant to be an informal and easy-to-use forum for exchanging thoughts and questions and for messages and news and pointers not always suitable for the more formal course website. Please very free to join the group, students and others. -- Nils Lid Hjort

(And no, it's not mandatory, and you can be a very good student without taking part in this group!)

Mon Aug 27 I discussed various matters from Chs 1, 2, and I used a bit of time to go through the Nils Exercise 1 on Old Egypt. Topics included hazard rate modelling, the empirical survival function, fitting of parametric models using maximum likelihood (ML), programming the log-likelihood function -- and then the eternal golden quadruple: counting process, at-risk-process, intensity process, martingale.

I've uploaded the R script com12a, which does various things for the Old Egypt dataset, including parametric ML fitting to the gamma and Weibull distributions. Do source("com12a") to run all of it, but also study its parts, and be ready to edit, to copy, to amend, to fix.

Next week I start Ch 3, with the Nelson-Aalen and Kaplan-Meier estimators etc.

I'm in the process of writing up a couple of exercises for the Nils Exercises collection. One of these is for next Monday, where the essence is as follows. Go to the data for the course and download "iud-data", w...

Mon Aug 20 I gave a broad introduction to the course and its themes, along with some practical information. Next week I continue with material from the ABG book's Chapters 1-2.

For Mon Aug 27, first do exercises 1, 2, 3, 4 from the books's Chapter 1. Then do the (somewhat long and perhaps somewhat laborious, but fruitful and useful) Nils Exercise 1, from the collection I just started (found at the course webpage).

Nils Lid Hjort

Welcome, everyone, to the stk 4080 / 9080 course, autumn semester 2018. Teaching takes place Mondays 12-15, in seminar room 123 at VB. The course material will be based on the Aalen, Borgan, Gjessing (2008, Springer) book, "Survival and Event History Analysis: A Process Point of View". In addition I will write up some exercises and lecture notes along the way.

Nils Lid Hjort