I am doing loglinear analysis answering the research question on 'What factors influence the satisfaction with the overal situation in the country?. I have 261 missing cases in my xtab. Would you suggest to work with such data or should I work with other variables?
Hi Lyubov
to respond, I need to know, how you calculated that 1000 sample?
I am sorry I did not undertand what do you mean by saying I found? I have dataset of 1000 cases, bu in some questions some people chose answer "I do not know" or refused to answer at all. In some other questions other people also chose such answers. In sum missing values resulted in 261, so I deleted those cases. But I think maybe I should work with other variables where there are less missing reponses to related questions.
the size of a sample to be representative of the population, must be calculated and collected randomly.
you talk of a sample of 1000. then how it was calculated that sample?
As you wrote, I understand that more than a "statistical sample" corresponds to a collection of cases.
Multi-stage cluster sample stratified by Russia’s eight regions
(excluding a few remote areas in the northern and eastern parts of
the country and Chechnya) and proportional to population size
and urban/rural population. Sample size: 1,000
Margin of Error: ±4.0 percentage points
Representative: Adult population
1000 cases is a large size sample. and 261 missing cases is too much. but I have some doubts about the variable used to calculate the variance, which is used to calculate the sample.
By my calculations, your sampling error is very close to zero, or very large variance.
but approximately 700 cases, it is still a good number.
Way too much. Missing data should be for at most 10% of the cases. If this happens you will need to determine whether you have a representative sample. This is determined with the optimal sample size formula for the entire population or, for the basic subdivisions of the population (major age groups, genders, main regions, etc.
Hello Lyubov Sartakova,
It is probably common to have a lot of missing data. On the favourite links page of my website there is an electronic stats book and from recall it has a section on how to treat missing data - casewise versus pairwise deletion from recall I could be wrong but it maybe useful for you.
Just as an example and it is not research just a fun pilot trial at the school where my daughter goes but you can read on the intouch page of my website Nov 12, the pedometer study 27 students given pedometers and booklets how many do you think returned the booklets just 6 students - God only knows what happened to the rest of these - missing !!!!
Good luck with the analysis. Thanks
Deborah Hilton Statistics Online
http://sites.google.com/site/deborahhilton/
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size