[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
"James W. Shaw" <[email protected]> |

To |
<[email protected]> |

Subject |
st: Difference between svylogit vs. svytab |

Date |
Mon, 13 Oct 2003 11:12:18 -0400 |

Dear Statalisters: I have data from a cross-sectional survey with 4,000 observations. The sample may be divided into two groups: group 1 (n=3,700) and group 2 (n=300). I wish to compare group 1 with group 2 on important demographic characteristics (e.g., age, gender, race/ethnicity, income). Several of the variables I wish to compare between group 1 and group 2 are categorical. For example, income is measured using 6 categories. There are missing data for several of the demographic variables. For example, out of 4,000 possible observations, I have 3,450 observations on income. I would like to use multiple imputation (as implemented by the -hotdeck- procedure) to impute values for missing data so that I can compare group 1 with group 2. Typically, the model test for the following logistic regression xi: logistic group i.income would yield the same result (p-value) as tab group income However, I am working with survey data. I have noted that xi: svylogit group i.income does not yield the same result (p-value) as svytab group income My questions are as follows: 1) Why do svylogit and svytab yield different results? 2) What does this difference mean? In other words, can I still interpret the model test for svylogit as a valid test for whether income differs between the two groups (or is associated with group membership)? 3) Assuming that I can interpret the model test for svylogit as a test for whether income varies between the two groups, could I use -hotdeck- as follows to impute the missing data and interpret the overall F test as a test for whether income differs between the two samples: hotdeck income, by(sex race) imp(10) command(xi: logistic sample i.income [pw = weight], cluster(psu)) parms( Iincome_2 Iincome_3 Iincome_4 Iincome_5 Iincome_6) keep(_all) store 4) Is there any other way that one could hot deck a 2-by-k (in the above example, 2-by-6) chi-square test using Stata? Thank you very much. Best Regards, Jim Shaw Research Associate College of Pharmacy The University of Arizona * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

- Prev by Date:
**RE: st: RE: bug un makematrix ?** - Next by Date:
**st: partial linear model** - Previous by thread:
**st: bug un makematrix ?** - Next by thread:
**st: partial linear model** - Index(es):

© Copyright 1996–2024 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |