Question:
A prof was delighted to be able to download the Census 2016 PUMF.
However, the weights do not seem to make sense to her.
These are her comments:
"I’m working with the PUMF file now and the weights look VERY odd. For starters, the file only contains 16 replicate weights – normally Stat Can provides 500 to 1000. For another thing, the individual (frequency weight) weight [variable name – WEIGHT] is nearly identical for all respondents. Is that what Stat Can intended?
Any information you can pass on would be really helpful."
Answer:
Subject matter has responded with the following:
“I’m happy to hear the prof is enjoying the PUMF.
We appreciate the concerns expressed by the user and we can assure them that the contents of the product are what Statistics Canada intended.
I would recommend that the users consult the PUMF user guide and codebook for answers to their questions. Chapter 3 – Sampling method, estimation and data quality should resolve any questions with regard to the Weight and replicate weight variables.”
Follow-up Question:
Here is the reply I received from my prof – and the answer provided by Subject division does not answer her question.
"Here’s an illustrative example where the application of the Weight variable creates some problems:
There are approximately 390,000 live births in Canada each year. When I run a cross tab to look at the number of women who are parents (couples or single parents) with a child under 1 year, I get a result of over 708,000, which just can’t be right."
I asked her for more detail to try to replicate how she got her answer and this is her reply:
"…my tabulation is restricted to women (Sex), of reproductive age (20 to 44 years per AGEGRP) who are in a couple with children or are single parents (CFSTAT) and who live with a child age 1 or younger (PKID0_1).
Here’s my Stata output for the derived variable (as per above) “newmom”
. tabulate newmom [fweight = rndweight]
newmom | Freq. Percent Cum.
------------+-----------------------------------
1 | 708,291 100.00 100.00
------------+-----------------------------------
Total | 708,291 100.00"
Follow-up Answer:
I received the following response from subject matter:
“On page 109 of the 2016 Census Individuals PUMF User Guide, the PKID0_1 variable includes children aged 0 or 1. Therefore it provides data on presence of children as of Census Day and it includes children that are under the age of 2.
As a result, the counts of children in Census family here presents higher numbers. Foster children are not included in this grouping.”
Thursday, February 21, 2019
Research Data Management Resources
Portage has recently published some new Training Aids for Research Data Management.
They include the following:
These bilingual resources are available for everyone here:
https://portagenetwork.ca/training-resources/portage-training-resources/
They include the following:
- an online training module - Research Data Management (RDM) 101
- Additions to the “Good Enough” series (1 page guides for busy people)
- Data Curation
- Dataverse
- FRDR
These bilingual resources are available for everyone here:
https://portagenetwork.ca/training-resources/portage-training-resources/
2006 Census PUMF
Question:
I was going through the user guide for the Census and I was a bit unclear whether I could calculate 95% confidence intervals (e.g. poisson approximation or gamma-distribution) on the age-standardized weighted proportions (of visible minority groups) simply through my direct age-standardization algorithm or if there are other considerations (e.g. variables WT1 through WT8).
Answer:
We received the following response from subject matter:
Poisson, gamma or other distributional assumptions are typically used for model based inference from data that is not sourced from surveys. It assumes that the event of interest is a random process following the hypothesized distribution and that your data is one realization from this random process. This assumption gives you a tool to derive a model based variance and confidence intervals. It is preferable to use a design-based method in the context of complex survey design data where the characteristic(event) is fixed for all units in the population of interest but the random error between the sample estimate and the true population parameter is due to the random sampling.
For standardized estimates from complex surveys, you need to account for the survey design information in the estimation of the variance. For the 2006 PUMF data file, the steps to do so are described in Chapter 3, section C.2 "Estimation of the sampling variability".
Specifically for the 2006 Census PUMF, you would compute 8 age-standardized estimates, once using each weight variable (step 1) and continue through to step 6.
The method depicted there provides you with 95% Wald confidence interval. It's the method most commonly used for convenience but it rests upon some assumptions, namely that the distribution of the estimator is normal which tends to be inappropriate for very small proportions on smaller domains of estimation. I assume it is less likely to affect you since standardization generally commends larger estimation domains.
I was going through the user guide for the Census and I was a bit unclear whether I could calculate 95% confidence intervals (e.g. poisson approximation or gamma-distribution) on the age-standardized weighted proportions (of visible minority groups) simply through my direct age-standardization algorithm or if there are other considerations (e.g. variables WT1 through WT8).
Answer:
We received the following response from subject matter:
Poisson, gamma or other distributional assumptions are typically used for model based inference from data that is not sourced from surveys. It assumes that the event of interest is a random process following the hypothesized distribution and that your data is one realization from this random process. This assumption gives you a tool to derive a model based variance and confidence intervals. It is preferable to use a design-based method in the context of complex survey design data where the characteristic(event) is fixed for all units in the population of interest but the random error between the sample estimate and the true population parameter is due to the random sampling.
For standardized estimates from complex surveys, you need to account for the survey design information in the estimation of the variance. For the 2006 PUMF data file, the steps to do so are described in Chapter 3, section C.2 "Estimation of the sampling variability".
Specifically for the 2006 Census PUMF, you would compute 8 age-standardized estimates, once using each weight variable (step 1) and continue through to step 6.
The method depicted there provides you with 95% Wald confidence interval. It's the method most commonly used for convenience but it rests upon some assumptions, namely that the distribution of the estimator is normal which tends to be inappropriate for very small proportions on smaller domains of estimation. I assume it is less likely to affect you since standardization generally commends larger estimation domains.
Wednesday, February 20, 2019
PCCF+ ID Variable
Question:
I’m trying to set up a PCCF+ file for someone and wanted to check…
My input file is a dataset with two columns: ID and Postal Code. My question is about the ID variable: are these randomly generated or is the ID pulled from an actual place?
Answer:
The ID needed can be from the researcher’s dataset or simply 1,2,3,…. that you put in to run the dataset. When I am helping researchers with the PCCF+, I suggest that they use the ID field from the dataset that the postal codes are from. That way, when they get the output file it is easier to match up to their dataset.
The rule for ID in the PCCF+ is that it is unique and up to 15 characters. I have used IDs containing letters and it has worked.
I have attached a PCCF+ Guide that Jeff Moon wrote that is very helpful in running the PCCF+. The example uses version 6B1 but process is the same regardless of the PCCF+ version.
** The original email had the PCCF+ Guide attached.
I’m trying to set up a PCCF+ file for someone and wanted to check…
My input file is a dataset with two columns: ID and Postal Code. My question is about the ID variable: are these randomly generated or is the ID pulled from an actual place?
Answer:
The ID needed can be from the researcher’s dataset or simply 1,2,3,…. that you put in to run the dataset. When I am helping researchers with the PCCF+, I suggest that they use the ID field from the dataset that the postal codes are from. That way, when they get the output file it is easier to match up to their dataset.
The rule for ID in the PCCF+ is that it is unique and up to 15 characters. I have used IDs containing letters and it has worked.
I have attached a PCCF+ Guide that Jeff Moon wrote that is very helpful in running the PCCF+. The example uses version 6B1 but process is the same regardless of the PCCF+ version.
** The original email had the PCCF+ Guide attached.
Census Agriculture Region (CAR) Boundary Files
Question:
I’m trying to locate ag region boundary files prior to 2001. I’ve looked through the various subdirectories on the EFT server by cannot find what I’m looking for.
Answer:
Looks like we have 1996 CAR boundaries in the GeoPortal http://geo.scholarsportal.info/#r/details/_uri@=3577293255$DLI_1996_Census_CBF_Eng_Nat_car
And something called the Agricultural Ecumene layers from earlier years (1986 and 1991)
If you load the layers to the map and go to the ‘Download’ tab and choose ‘Download all’ we’ve packaged the original data from StatCan in there (should be in E00 format). For 1996 this file was named ‘car96cangeo.e00’.
I’m trying to locate ag region boundary files prior to 2001. I’ve looked through the various subdirectories on the EFT server by cannot find what I’m looking for.
Answer:
Looks like we have 1996 CAR boundaries in the GeoPortal http://geo.scholarsportal.info/#r/details/_uri@=3577293255$DLI_1996_Census_CBF_Eng_Nat_car
And something called the Agricultural Ecumene layers from earlier years (1986 and 1991)
If you load the layers to the map and go to the ‘Download’ tab and choose ‘Download all’ we’ve packaged the original data from StatCan in there (should be in E00 format). For 1996 this file was named ‘car96cangeo.e00’.
Tuesday, February 19, 2019
Annual Labour Force Survey (LFS) Files
Question:
Some students are looking for annual LFS files instead of monthly. Are those available somewhere? Are the old annual Equinox files archived and available?
Answer:
We only have the monthly LFS PUMFs available at this point in time, but have been exploring options for annual files as well. I can’t speak to the Equinox files however though.
Some students are looking for annual LFS files instead of monthly. Are those available somewhere? Are the old annual Equinox files archived and available?
Answer:
We only have the monthly LFS PUMFs available at this point in time, but have been exploring options for annual files as well. I can’t speak to the Equinox files however though.
Thursday, February 14, 2019
Ethnicity and Census
Question:
I have received what looks like a deceptively simple request. One of our researchers is looking for ethnicity and/or country of origin by CT (although he would take DA, obviously) for the 2016 census. I can’t seem to find anything on the Statcan website, and the FTP site doesn’t seem to have anything like the cumulative profiles at the DA level as from previous years.
Answer from DLI List:
Have you tried the Canadian Census Analyzer from CHASS (http://dc.chass.utoronto.ca/census/)? It has ethnic origin and place of birth for both CT and DA. You do have to be a subscribing institution to access it but the website says UBC is.
Census subject matter has responded with the following:
"We only have a limited number of standard products available at these smaller geographies. This would need to be done as a custom request."
I have received what looks like a deceptively simple request. One of our researchers is looking for ethnicity and/or country of origin by CT (although he would take DA, obviously) for the 2016 census. I can’t seem to find anything on the Statcan website, and the FTP site doesn’t seem to have anything like the cumulative profiles at the DA level as from previous years.
Answer from DLI List:
Have you tried the Canadian Census Analyzer from CHASS (http://dc.chass.utoronto.ca/census/)? It has ethnic origin and place of birth for both CT and DA. You do have to be a subscribing institution to access it but the website says UBC is.
Census subject matter has responded with the following:
"We only have a limited number of standard products available at these smaller geographies. This would need to be done as a custom request."
Subscribe to:
Posts (Atom)