Discussão:Pesquisa:Inquérito crack Brasil/Questionário Sexo-idade Mortalidade

Da wiki Wikiversidade
Ir para: navegação, pesquisa

Recursos (resources)[editar código-fonte]

Model questionnaires[editar código-fonte]

Hi all,

Previously, I had sent along the sibling survival questions from a DHS survey that was conducted in Mozambique. Recently, I also found sibling survival questions that were asked in Brazil, in the 1996 DHS survey. I'm not sure how different Portuguese is in Brazil and Mozambique, but it looks like at least a little of the phrasing is not the same. I'm sending the Brazilian and Mozambican versions of the questions in a separate email. I think the questions currently in the survey are based on the Mozambique ones, but perhaps the Brazilian ones would be a better model -- what do you guys think?

Hope all is well, --Dfeehan (forgot to sign --Solstag 23h01min de 5 de Dezembro de 2010 (UTC))

Q 8i (tie strength to sibling)[editar código-fonte]

The tie strength question with the sibling is based on the ideas developed in a paper by Marsden and Campbell (http://dl.dropbox.com/u/1437602/measuring_tie_stength.pdf). --Msalganik 23h51min de 20 de Dezembro de 2010 (UTC)

Questões abertas (open questions)[editar código-fonte]

Block order[editar código-fonte]

Hi, is there some reason for the blocks to be disposed as they are? I just changed the order of the migration and mortality blocks. The reason is I thought it wasn't very good to ask the person about international money transfers right before you ask about their drug history (the following block is personal drug use), and because it might be good to oscillate the tone of the conversation (so we have siblings using drugs -> migration instead of siblings using drugs -> death). What do you guys think? I tried to contact you instantly to decide on this before the next day, but I couldn't, so I'm sending the questionnaire anyway and we can revert this later. Hugs, --Solstag 20h57min de 7 de Fevereiro de 2012 (UTC)

I'm not sure if there is any particular reason for the order. I guess it makes sense to put them in the order that you think is most natural. This is something that we can figure out more in the pilot testing. --Msalganik 00h24min de 8 de Fevereiro de 2012 (UTC)

Questões resolvidas (resolved issues)[editar código-fonte]

More timing[editar código-fonte]

I don't understand when we ask in [AGESEX-SIBq] whether they "lived here in the 12 months before"; is it more like "for the whole duration of the 12 months before" (makes some sense to me, but still dubious) or "at any moment during the 12 months before" (makes less sense, but still might be), or "at the exact moment 12 months before" (doesn't make much sense to me)? --Solstag 20h00min de 6 de Fevereiro de 2012 (UTC)

Ok, I think I get it, it's probably "for any period during the 12 months before", since we actually wanna know if he counts as "migrated in the last 12 months". Right? --Solstag 10h16min de 7 de Fevereiro de 2012 (UTC)
Yes, you are right: we want to know if he counts as having arrived or having moved away from the city within the last 12 months. So if the sibling does live in the city now, we want to check if s/he has been in the city for at least a year, or if, instead, s/he migrated into the city within the past year. Conversely, if s/he doesn't live in the city now, we want to check if s/he did live in the city a year ago to see whether or not s/she just migrated out. ( We are counting migrants here, but not migration events, since people could, in principle, have moved more than once over the course of a year.) --Dfeehan 14h58min de 7 de Fevereiro de 2012 (UTC)
After a somewhat long discussion with Dennis, we managed to understand what was important and agree to a solution to this, which involved considerably changing many of the questions involved and creating other ones. --Solstag 17h13min de 7 de Fevereiro de 2012 (UTC)

Info on emigrants[editar código-fonte]

For people who are reported to have moved away, we think it would be interesting to collect information on the education level:

has NOME completed ensino fundamental?

and also a question on remittances :

Do you think NOME regularly sends money back to Brazil?

These are both important to the study of migration because many countries are interested in who emigrates (for example, is it manual laborers or doctors?) and also in how people who have migrated might contribute to development in the local economy (through remittances, ie, sending money back)

Hm... study question ok, but why "think" in money question? Any reason not to ask directly, I mean, under the "best guess" assumption? --Solstag 21h48min de 3 de Fevereiro de 2012 (UTC)
I agree with you that we want the "best guess" assumption for this question; that's why I used the word 'think' -- I wanted to imply that it was OK for the respondent to give us her best guess, even if she didn't know for certain. There's probably a better way to phrase that, though, both in English and in Portuguese! It sounds like we are on the same page in terms of what we want to ask, though, so however it makes sense to ask that in Portuguese is what I think we should use... -- 19h38min de 6 de Fevereiro de 2012 (UTC)

Implemented. --Solstag 10h57min de 7 de Fevereiro de 2012 (UTC)

Timing questions[editar código-fonte]

We think it would be useful to add a question about how long ago the event happened in the sibling, mortality and migration sections. So, the first question we ask about every death would be 'how many months ago did the death happen?'; similarly, the first questions for the migration section would be 'how many months ago did the person move?' This will (a) help us assess how much people's ability to recall events about their alters diminishes over time; and (b) help us look at seasonal patterns in these events.

Hm.. ok, I'll find a way to ask that. --Solstag 21h46min de 3 de Fevereiro de 2012 (UTC)
Done! --Solstag 10h12min de 7 de Fevereiro de 2012 (UTC)

Don't know[editar código-fonte]

For the questions about crack use, for example 8j0, it seems reasonable that some respondents may not definitively know whether or not their siblings have ever used crack, but that they might have a good guess. In that situation, it seems like we should either have the interviewers clarify that we want the respondent's best guess, or we should have a "don't know" response category in addition to "yes" and "no". Matt and I think it probably makes more sense to have the respondent give her best guess -- what does everyone else think? [Dennis 27 Jan 2011]

I agree with "best guess", otherwise people may "escape" towards "don't know". That should be a general instruction to the interviewers, but we could remind them at those specific questions. --Solstag 21h51min de 3 de Fevereiro de 2012 (UTC)
Ok, I left a generic instruction at the beginning of the questionnaire. --Solstag 08h54min de 7 de Fevereiro de 2012 (UTC)

Duplicate question[editar código-fonte]

Hi everyone,

We realized that we seem to be asking how many people the respondent knows who used crack/similares in the past 6 months twice: once in the NSUM section, and one at the start of the age-sex module. Our suggestion is to change the question at the start of the age-sex module so that it is something the interviewer fills out based on the previous responses, rather than something that gets asked of the respondent.

What do you all think?

--Dfeehan 15h12min de 3 de Fevereiro de 2012 (UTC) and Matt

Alright, if we don't want to cross check, which perhaps we don't since the questions are asked so close to each other, we can have it like that : ) --Solstag 21h43min de 3 de Fevereiro de 2012 (UTC)
I left a note for the interviewer, so he remebers to recover the previous answer. --Solstag 21h07min de 6 de Fevereiro de 2012 (UTC)

Tie definitions for age-sex sections[editar código-fonte]

We decided to use contacts over the past 12 months (instead of 2 years) for the definition of a tie ('to know') in the study, so we should update the definitions of a tie used in the age-sex sections to cover 12 months instead of 2 years as well.

Also, in the mortality section, I think there may be a typo in the translation -- we want to ask about deaths that happened in the past 12 months, not the past 12 years.

And, finally, in the introduction to each question, it seems like it would make sense to remind people that we are referring to residents of their municipio. So, when we remind people about the definition of 'to know' before the mortality and migration questions, we suggest adding a third bullet point that says 'people who are residents of this municipio'

Will fix --Solstag 21h44min de 3 de Fevereiro de 2012 (UTC)

Evolution question for siblings[editar código-fonte]

Hi everyone,

We think that it makes sense to change the order of the sibling questions that we will use to produce estimates of the evolution of the crack epidemic over time. Currently, we ask:

[AGESEX-SIBl] NOME usou crack e similares alguma vez na vida? — Por "crack e similares" entenda-se: crack, pasta base, merla ou oxi, fumados em cachimbos, copos ou latas.

[AGESEX-SIBm] [só VIVO] NOME usou crack e similares pelo menos uma vez nos últimos 6 meses?

[AGESEX-SIBn] [pular se AGESEX-SIBm=NÃO] Quando NOME usou crack e similares pela primeira vez?

[AGESEX-SIBo] [pular se AGESEX-SIBm=NÃO] Quando NOME usou crack e similares pela última vez?

It seems to us that it makes more sense to ask [AGESEX-SIBl], and *then* [AGESEX-SIBn] for everyone who answers yes to [AGESEX-SIBl]. In other words, for every sibling who is reported to have used crack, we should ask when s/he started using crack. Then we ask about the last 6 months [AGESEX-SIBm], and if the sibling hasn't used crack in the last six months, we ask when s/he stopped (AGESEX-SIBo).

What do you think?

--Dfeehan 16h36min de 3 de Fevereiro de 2012 (UTC) and Matt

Perfect, will change that. --Solstag 21h49min de 3 de Fevereiro de 2012 (UTC)

New question about murder as the cause of death?[editar código-fonte]

We thought it would be interesting to consider asking if any of the siblings reported dead was murdered. So, after question 8h, asking

  • 8h1 was he murdered?

(That may not be the best way to word the question.) We could also ask, after the scale-up mortality questions,

  • 14 Out of all of the people you know who died in the last 12 months, how many of them were murdered?

This would serve as an additional validity check of the two methods, since we could compare the estimates to the official ones that are available for each city. It would also then let us study an interesting set of questions around who knows people who are murdered, and who knows people who use crack. The difference in responses to these questions across cities could be also be very interesting. There's lots of things we could look at. For instance, my understanding is that, very roughly, Sao Paolo is quite divided into rich and poor areas, while Rio is more mixed together. It would be interesting if we could see that the sort of people who have their lives disrupted by murder or drug use depends on how the city is structured; it might be that a range of richer and poorer people are affected by crack and murder in Rio, while only the poorer people are affected in Sao Paolo. [Dennis 05 Jan 2011]

Not possible. Crime data is really bad and no basis for comparison. And we can't afford paying life insurance to our interviewers.
--Solstag 13h27min de 26 de Janeiro de 2011 (UTC)
Chico forgot to include money in the grant for life insurance? What's going on? :) More seriously, we see your point. --Msalganik 17h54min de 27 de Janeiro de 2011 (UTC)

New questions about evolution data[editar código-fonte]

In the sibling questionnaire, we want to change the simple question about crack use to something more complex that will give us data about the evolution of crack use:

  1. did he use at least once in the last 6 months (if YES go to 3)
  2. has he ever used crack? (if NO finish)
  3. when did he first use crack?
  4. when did he last use crack?

--Solstag 08h29min de 20 de Dezembro de 2010 (UTC)

Done. --Solstag 13h29min de 26 de Janeiro de 2011 (UTC)

Q 2.[editar código-fonte]

Acho melhor repetir da mesma forma. Ganhamos em consistência! --Neilane

Não entendi bem, da mesma forma que qual? --Solstag 06h45min de 23 de Novembro de 2010 (UTC)
Fechando por passar o prazo de validade. hehehe --Solstag 13h52min de 26 de Janeiro de 2011 (UTC)

Place of life and death[editar código-fonte]

About the sibling questions, we need to get more information about where the sibling is living or died. The reason is that we need to be able to assign that siblings death and/or crack use to the right city. For example, if you have a sibling that uses crack and lives in Manaus, we want to make sure that person gets counted in Manaus. Right now, we were only recording if they live in the same city as the respondent. So, in each "branch" of the survey---the branch if the sibling is alive and the branch if the sibling is dead---we want to collect the city and state where they live (or did live). I am not sure the best way to ask this in Portuguese. Also, we realize that collecting this free text information will be difficult to code. Perhaps we can list the 27 cities where we are doing the study, which the interviewer can circle, and we can take a free-response for the others. Since most of the siblings will probably live in the 27 cities that will save a lot of work. --Msalganik (by e-mail --Solstag 23h14min de 5 de Dezembro de 2010 (UTC))

This is going to take some time from the survey, which is already in danger of exceeding the contractual time. But more importantly, I see no reason why most of the siblings who do not live in the same city are going to live in one of the 27 other cities. Perhaps Dimitri can give us a better picture of this. We might end up with a record of cities for which we have no other information. Is that also useful? --Solstag 00h47min de 6 de Dezembro de 2010 (UTC)
I thought that these 27 cities are some of the biggest cities in Brazil. In the US, I think a non-trivial portion of our population live in the 27 largest cities. However, for our purposes what we really need to know is if they are in one of the 27 cities or somewhere else (in other words, 28 categories). Anything that we could get about the location of people not in the 27 cities would be a bonus. --Msalganik 03h00min de 6 de Dezembro de 2010 (UTC)
Hi! I understand the utility of this question. But there it goes some additional demographic information about Brazil which can help us to choose the best formulation question and wording. Those 27 cities listed are not all the biggest ones (indeed, we have some capitals of provinces that are far below 500 thousand inhabitants). Anyway, regionally all of them are the most prominent and densely populated and also concentrate the public health services. So, it is much more likely that people who does not live in the capitals will come to die in there. In contrast, usually, residents in the more urbanized cities (all 27 capitals) will probably come to die in the hospitals of their city of residence - once I heard about some statistics regarding IDUs' deaths in Belo Horizonte, and the majority of them were established residents in the capital (no statistics otherwise). However, the migratory status of siblings does not depends on "kinship selectivity". That is, I believe addicted siblings of respondents have the same chance of living away as other Brazilians do (and this is a very small proportion of the general population). In conclusion, I would say that if we could include the question about "location of residence" of sibling crack users (and not "death location"), we should keep it (but again conditioned on time and questionnaire resources). [Dimitri 06DEC2010]

I just want to clarify some things... as far as I understand, this is about the sibling section only, not about the crack or mortality sections. Which means this is about question Q 8k, which currently reads "Does NAME lives in the same city as you do?".

You want it to read like "Where does your sibling live?". However, making that an open-ended question takes up more time, so there should be a good reason for us to do this. You also said that it is only useful if the city is one of the 27 cities we know of. Truth is, I have no idea exactly why you are asking this question.

The only thing I can see this being useful for is to tell whether the statistics differ between people who have distant siblings and those who haven't, or whether it differs between the people who report having distant siblings and the siblings reported by them.

Personally, I don't see why the first thing would be interesting. The second thing is more like a methodological issue and looks more promising; if we find out that self-reported crack use is lower than sibling-reported for people in the same city, we get a city-specific ratio for this under-reporting.

Ok, I wrote a lot already, perhaps I'm missing the point. Hopefully I missed it by coming up with something at least worth reading :P

--Solstag 05h40min de 8 de Dezembro de 2010 (UTC)

OK, I see from Ale and Dimitri's comments that this is a bit confusing. We want to know where the sibling lives (or lived if they are dead) so that we can assign their crack use or death to some city. Let's say we want to estimate the crack use rate in Sao Paulo using this sibling data. One approach is to take all the siblings reported by respondents from Sao Paulo who use crack (say 500) and divide by the total number of siblings reported respondents living in Sao Paulo (say 50,000). In this case we would estimate that 500/50,000 or 1% of the people in Sao Paulo use crack. However, there is a problem here because we are not sure that these siblings actually live in Sao Paulo and there might be siblings of people in Rio who live in Sao Paulo. Therefore, what we want to do is take the number of siblings reported living in Sao Paulo (from people in all 27 cities) and then divide by the total number of siblings living in Sao Paulo (from people in all 27 cities). If we wanted to make national level estimates about crack use, then we would not need to collect the city in which the sibling live. However, we want to estimate crack use at the city level so we need to know in which city to put each of these crack users (and non-crack users). The same argument as above applies to estimate mortality at the city-level. I hope that is more clear. Let me know if you have any questions. --Msalganik 03h12min de 10 de Dezembro de 2010 (UTC)

In a meeting with neilane, chico and dimitri, we all agreed that we don't think it is worth taking the time to ask about the location of siblings who do not live in the same city. The reason is that the fraction of the population that would live in another of the capital cities is expected to be negligible - local migration dominates - and the answers expected to be of very low quality - exact locations not known. --Solstag 18h32min de 17 de Dezembro de 2010 (UTC)
OK, Ale, Dennis, and I talked about this some on Skype and then Dennis and I had some time to think about it. We understand that the data might not be very accurate if, for example, Ale's sibling moved to Manaus, he might not know if she lived in Manaus or the suburbs of Manaus. But, if we try to collect this data (again, we would need once question with 28 answers, the 27 cities+ other are the only options), then we could calculate the estimates two ways: 1) using only reports of siblings in the same city and 2) using all siblings. If we see which produces better mortality estimates then we have learned something. Although, we want to be clear about our preferred estimator /before/ we collect the data. Imagine for example that we do the study (without collecting the exact location of the siblings) and our sibship estimates are off. We might always wonder if the sibling that moved might have explained it. So that's our thinking about this now, but I'm sure this conversation will continue. --Msalganik 20h51min de 27 de Dezembro de 2010 (UTC)
Ok, I'm reopening this for the last time. Recalling what I explained to you guys, the issue is not only with accuracy, but with statistical significance. Since we assume, and we have good reason for that, that the amount of data is going to be statistically insignificant, we are not going to be wondering if the sibling that moved might have explained it: we know beforehand that it could not. As Dimitri explained to me, and I to you, Brazilians migrate very locally and only travel long distances upon occasional economic drives and, even then, frequently ending up in satellite cities. But since you guys do not seem convinced and think this is such a big deal, I am going to raise this again with Dimitri, Chico and Neilane. --Solstag 06h47min de 6 de Janeiro de 2011 (UTC)

We have concluded that this is really really really not viable and not a good idea in our context, for the reasons outlined before. --Solstag 13h40min de 26 de Janeiro de 2011 (UTC)

When we are collecting city-state of the siblings who have died, do we want place of death or place of residence when they died. For example, I could live in Princeton, but die in New York.[editar código-fonte]

For the sibling questions, and for the siblings that died, we want to find out where the home of the sibling was before they died - city and state. The complication is that, for example, I might live in Princeton, but go the hospital and die in New York City. For the purposes our estimate, we want to know that I used to live in Princeton, not that I died in New York. Dfeehan said that one way to think about this is where would they have lived if they didn't die, but that is a bit confusing too.

I guess in English the best way to ask this might be: "Where did [sibling name] live before he/she died? If they moved to a medical center or hospital for treatment, please consider only the place of their last home before moving to the medical center." Name of city: Name of state:

Is there a concept for this in Portuguese? Do you have an idea for how to phrase this question? Also, I think it is OK if it is a little long because it only comes up in the mortality branch of the sibling questions so it will not be asked to every respondent, every time.


--Msalganik (by email --Solstag 23h14min de 5 de Dezembro de 2010 (UTC))

Ok, in case we go forward with this, it is easy, we just ask where the person had his last fixed residency ("residência fixa") before his death. --Solstag 00h47min de 6 de Dezembro de 2010 (UTC)

Time windows[editar código-fonte]

Q 8e[editar código-fonte]

Na questão 8e, vamos mudar para 6 meses para padronizar ou necessita ser 12 meses? --Neilane

Bem, nossa definição de conhece ainda quer bater com a do RDS? Se sim, acho que era 12 meses, né? Afinal, nessas definições é possível haver irmãos que não se conhecem! --Solstag 06h44min de 23 de Novembro de 2010 (UTC)
Neilane, entendi melhor agora e não sei se concordo. No questionário scaleup a definição de conhece usava dois anos e não 12 meses. Mas, de fato, essa questão tem um propósito diferente, que eu não consigo entender. Ou talvez nem isso, ela parece desnecessária para os nossos objetivos.--Solstag 22h09min de 6 de Dezembro de 2010 (UTC)
Eu, dimitri e chico conluímos que essa questão não tem relação com a definição de conhece. Passamos para 6 meses pois assim podemos possivelmente corerlacionar com a forma como a pessoa obteve informação sobre o uso de drogas. --Solstag 18h36min de 17 de Dezembro de 2010 (UTC)

Q 9 (mortality question)[editar código-fonte]

For the how many people do you know that died questions, we want to change the time window to 12 months, rather than 6 months.

--Msalganik 20h48min de 6 de Dezembro de 2010 (UTC)

Is there any reason not to use the same time window that our definition of drug use does? I guess it is ok since they seem independent, but I'm curious why 12 months?
Also, Neilane made a similar suggestion for Q8e, which asks whether the person has seen a living sibling in the last 12 months. But she wanted to make it 6 months from 12, to fit the drug use time window. I don't see where Q8e is needed in computing any statistics, so perhaps it could be dropped. But in case we want to keep it, the length of time seems arbitrary. Any suggestions?
--Solstag 22h09min de 6 de Dezembro de 2010 (UTC)
We want 12 months because mortality data is reported over a 12 month period. If we estimate over a 6 month period we might run into a problem if there is seasonal variation (more people dying in the winter, for example). Further, if we ask about 12 months we should learn about twice as many dead people as if we ask about 6 months. The more dead people we know the age and sex of the better. --Msalganik 03h28min de 10 de Dezembro de 2010 (UTC)
Done. --Solstag 18h36min de 17 de Dezembro de 2010 (UTC)

Date of birth and date on official records[editar código-fonte]

When asking about the age of people known, do we want to ask about biological age of age in official documents? --Msalganik (by e-mail --Solstag 00h27min de 6 de Dezembro de 2010 (UTC))

This seems to depend on "dimensional analysis" of the equations if public records are used as input. If things are independent or there is no public record input, then we have to choose exactly what we want to measure. In principle, I think knowing about real date of birth might be more interesting. But it depends on how other surveys deal with this that we might want to compare ours to. Probably they don't, so we must "guess" whether people would answer with their official or natural age. I have no idea. Dimitri? Chico? Survey people? --Solstag 00h53min de 6 de Dezembro de 2010 (UTC)
We are trying to find out what data they have in the death certificates. If they have "biological age" (real age) that is what we will want. If they only have "recorded age" then we will have a more difficult decision.--Msalganik 03h02min de 6 de Dezembro de 2010 (UTC)
Neilane told me something very troubling yesterday: that DataSUS (The Unified Health System Data Agency) death records today contain only events from 2007 back. Since we are asking about people who died in the last 6 - or 12 - months, we might not be able to compare this data to public records at all; not for a very long time. Also, we don't need records about last residency and place of death, right? Cause that would come from the exact same database, though on the bright side it does incorporate the notion discussed in the other topic, of place of residency differing from place of death. --Solstag 06h02min de 8 de Dezembro de 2010 (UTC)
Yes, this is not great. But, Dennis and I talked about this and we can take the mortality data from 2000, 2001, 2002, . . . .2007 and project forward mortality for 2010. If mortality rates are pretty stable which I think they might be, then we are OK. It is just one more piece of "noise" in the process. Also, in 2014 (or whenever the 2010 data becomes available) we can always check out projection. --Msalganik 03h15min de 10 de Dezembro de 2010 (UTC)
Dimitri talked to some people and he believes the difference of biological and official age is a minor issue. --Solstag 18h32min de 17 de Dezembro de 2010 (UTC)

Exact ages or age categories[editar código-fonte]

Thanks for creating the wiki!

I had a couple of questions about this part of the questionnaire:

1) I think it would make more sense to collect exact ages, rather than asking people to put them into categories. We can analyze the data in categories later, if we decide that's the way to go. What do others think?

--Dfeehan 14h30min de 19 de Novembro de 2010 (UTC) (esqueceu-se de assinar --Solstag 03h08min de 23 de Novembro de 2010 (UTC))

It was suggested that people might take longer to answer if you ask them to give the exact age. As we are very worried about the length of the questionnaire, Neilane changed this. I'm not so sure it would take longer, this clearly deserves more thought.
--Solstag 06h39min de 23 de Novembro de 2010 (UTC)
It is much easier for us statistically if we can collect an exact age. Neilane thinks that this might take longer for respondents than using categories, but using categories will make the analysis *much* harder. Therefore, we would like to ask for the exact age, even if it might take respondents a bit longer. This change would affect both the crack questions (3 and 5) and the death questions (11 and 13) (http://pt.wikiversity.org/wiki/Pesquisa:Inqu%C3%A9rito_crack_Brasil/Question%C3%A1rio_Sexo-idade_Mortalidade). One open issue that Chico explained to us is that for some people the biological age is not the same as the age in official records. This is especially possible for older and rural Brazilians who were not registered when they were born. We will need to figure out if we want to ask about biological age or age in the records. I will add this to the list of open questions. --Msalganik (via e-mail --Solstag 22h53min de 5 de Dezembro de 2010 (UTC))
Yes, the benefits are evident and asking for exact ages won't take that much more time. I raised this issue of people having different birth dates in official records before. Good that chico brought that up again, I will start a different discussion topic to deal with that. --Solstag 22h53min de 5 de Dezembro de 2010 (UTC)

Once Neilane has a say about this I'll change the questionnaire to exact ages. If anyone has an idea about how to best formulate and word this question I'm all ears. --Solstag 22h55min de 5 de Dezembro de 2010 (UTC)

Ok, changed to exact ages. --Solstag 18h32min de 17 de Dezembro de 2010 (UTC)

Q 3. , 5. , 11. e 13.[editar código-fonte]

Quais faixas etárias usar? --Neilane

Não vamos mais usar faixas etárias. --Solstag 18h32min de 17 de Dezembro de 2010 (UTC)

Survey method[editar código-fonte]

We were originally talking about two different ways of asking about the age and sex of the crack users, and of deaths. I think the first way we proposed (more like a birth history -- method 1 in the "Proposed Additions to Drug-Use Survey" document that Msalganik and I sent around) is likely to give better results, but I'm not sure. The questions in the survey right now are method 2, which asks for the total numbers. The reason I think the first way is likely to work better is that it forces the respondents to think through each person they report about, one at a time. What do others think about this?

--Dfeehan 14h30min de 19 de Novembro de 2010 (UTC) (esqueceu-se de assinar --Solstag 03h08min de 23 de Novembro de 2010 (UTC))

The main concern here was the length of the interview. It must not go over 30 min 'cause that is what is being contracted. Thus, we might not be able to afford asking about each person individually. Neilane did some length testing on herself, so maybe she can share her view on this.
--Solstag 06h39min de 23 de Novembro de 2010 (UTC)
We have decided on the shorter method. --Solstag 18h32min de 17 de Dezembro de 2010 (UTC)

Brazil or Moçambique questionnaire[editar código-fonte]

Hi everyone,

Chico looked at those two questionnaires during his visit to Princeton and he prefers the questionnaire from Mozambique.

-Msalganik 15h22min de 4 de Dezembro de 2010 (UTC)

I don't think either are very good, we have already started with the Mozambique questionnaire, but we'll have to change a few things anyway. Neilane already raises a few issues above, but there are more. The wiki was created before the Brazil questionnaire showed up, so everything here relates to the Mozambique model. --Solstag 22h37min de 5 de Dezembro de 2010 (UTC)
OK, good news. Dennis and I thought the Brazilian questionnaire would be helpful, but I guess it turned out not to be. --Msalganik 03h04min de 6 de Dezembro de 2010 (UTC)