i have a data set in which there are about 13500 individuals distributed in 500 clusters, and also education level of individuals in and other variable. i want to generate a variable in which i want to measure education of These
cluster-level individuals education, defined as the percentage of individuals with secondary education in the cluster, and categorized as: low, middle, and high (cut-off at median value in all clusters combined; "middle" referring to the proportion at the median value, "low" referring to the proportion below the median value, and "high" referring to the proportion above the median value)
the variables in spss file are attached for you kind information please.
regards
atta
Hi,
Using the PPS sampling method.
http://www.rhrc.org/resources/general_fieldtools/toolkit/55b%20pps%20sampling%20technique.doc
http://www.who.int/tb/advisory_bodies/impact_measurement_taskforce/meetings/prevalence_survey/psws_probability_prop_size_bierrenbach.pdf
Hello,
You can compute your variable in STATA as follows:
First, generate a binary variable X (1=secondary completed, 0= not completed). Then use "collapse" command at cluster level with sub-command mean of variable X. Multiply the newly created variable of mean of X by 100, you will get the percentage of individual in a cluster with secondary completed education. Now, you can fined the median using "sum" command with option "detail". After getting the overall median of 500 cluster, you can classify the cluster level education as "low", "medium" and "high".
The syntax is,
gen secedu=0
replace secedu=1 if education>=6
collapse (mean) secedu, by(cluster)
replace mean(secedu)=mean(secedu)*100
sum mean(secedu),detail //you will find median
gen edulevel=.
replace edulevel=1 if mean(secedu)M (median)
To use collapse command correctly please see the help for the command.
I hope this will help.
You can compute them using SPSS please, go the compute put your new and old variables as you like and continue the process
You might consider modifying your a priori logic for categorizing your education achievement variable, since getting a good distribution around the median may be limited by those achieving the exact median value, instead of, say, an alternative range of values around the median, like the interquartile range between the upper and lower quartiles, for an inner 50% middle grouping. Then use those above and below the quartile values as high and low groupings. The point here is that you want to create groupings that allow for the maximum natural variation present in the distribution of the grouping variable to influence associations with other variables. Terciles could work just as well...upper, middle and lower thirds.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS