STAT 410 96-2 Assignment 10 Solutions
-
The systematic samples gave me the following table of sample means:
Sample | Mean Employment Income | Mean Government
Income |
1 | 176324.750 | 1241.625
|
2 | 160657.500 | 2855.625
|
3 | 147755.600 | 2631.000
|
4 | 150789.250 | 295.875
|
The grand means are just the column means and I got
the 2 Ybars to be $158881.7812 and $1756.0312. The variances asked for
in parts 1 and 3 are obtained by taking these numbers away from the
columns of sample means above, squaring, summing and dividing by 4.
I got the variances to be 124172826.8 and 1092836.70 leading to
standard errors of $11143.2862 and $1045.3883. The variances for
SRS are given by S^2(1-8/32)/32 and you have to compute the two population
variances. I get 2084275858 and 8855182.9 which lead to standard
errors of $13978.5858 and $911.1385 so that, as expected systematic
sampling is better for employment income (where there is a fairly
strong trend) and worse for government income (for which the ordering
seems more or less random).
- There are two ways to do this question. The variance of ybarbar from
a cluster sample of n clusters is S^2_M (1-n/128)/(nM^2) and we want to set
this equal to S^2(1-36/1024)/36 and solve for n after replacing S^2 and
S^2_M by estimates. The first and easiest way is to take your estimate
of S^2, that is s^2, from assignment 3 where you drew a SRS and your
estimate of S^2_M to be the sample variance of the 6 cluster totals
on the midterm. The second, and better way, is to follow the discussion
in class and use the raw data provided to fill in the analysis of variance
table for the data, then complete the estimated ANOVA for the population.
This method may be better because the desired value of n depends only
on the ratio of your estimates of S and S_M. The second method
uses a ratio estimate where the top and bottom are probably correlated
positively and so may well estimate n better even though the
srs estimate of S^2 itself is likely to be more accurate
than the cluster sample estimate of S^2 based on the ANOVA table; look
back at Question 6.4.
The questions.