Given what you know about the shape of some of these distributions,
recode the variables pblack, phisp and povpc into two or three
categories. You might need to play around with this, first looking at your
histograms and means and medians from last week, and perhaps discussing your
ideas decision with a lab partner.
HINT 1: Remember what you know about recoding from your exercise in
ArcGIS with net migration.
HINT 2: If you want to compare that hand counted table you produced for
homework 2, you might want to make the breaks the same as they were for those
data.
What
are your categories?
pblacnew
Low
_____ to _____ Medium _____
to _____ High _____ to _____
phispnew
Low
_____ to _____ Medium _____
to _____ High _____ to _____
povpcnew
Low
_____ to _____ Medium _____
to _____ High _____ to _____
Briefly
describe your logic of classification for each variable. Why did you break the
variable up this way?
Remember
to name your variables, provide sensible variable labels, and save the
variables to your data set because you will be using them in future analysis.
Please
attach frequency distributions of your new categorical variables to this lab.
Analyze
®
Descriptive Statistics ® Crosstabs
Now
you have recoded each of the variables pblack, phisp and povpv
into a smaller number of categories, you can look at the relationship between
ethnic composition and poverty more systematically. For example, you can ask,
is the proportion black associated with the percent in poverty across MA
census tracts?
Note,
that in this example, proportion black is the independent variable and
the percent in poverty is the dependent variable. Briefly explain why
race rather than poverty is the independent variable?
Using
the variables you re-classified into 2 or 3 categories above, construct two
tables examining
(1)
the
percent in poverty by percent black for all MA census tracts (note: you could
use your analysis here to compare to your findings from your hand count of the
5% sample for the same variable)
(2)
the
percent in poverty by percent hispanic for all MA census tracts.
Make sure you label the new variables and category values appropriately in Variable View so you (and we) can read your output.
Percentage the tables in the appropriate direction.
Describe the patterns you see in a few sentences. Is the pattern different for
blacks and hispanics/latinos?
Graphs
®
Interactive ®
Bar
Sometimes,
you may want to look at differences in some continuous value, like income for
different groups. So for example, you might want to know differences in mean
income between tracts with different ethnic compositions.
There
are several different ways to do this (using the Select command and
calculating means, medians etc for different subgroups of data like you did in
Lab 1, using the Compare Means command in the Frequencies menu).
In what follows, however, we will use the Graphs menu to make
informative Interactive Bar Graphs.
Construct
a bar graph that shows the mean difference in median family income (mdfam_in)
for tracts with different ethnic compositions (use your categorical black
or hispanic variable to make the comparison). Attach the graph to your lab.
Transform
® Compute ® meaned
COMPUTE meaned =
((less_g9*1)+(g9_g12*2)+(hs*3)+(so_col * 4)+(assoc*5)+(bach *6)+(grad*7))/
pop2000.
EXECUTE .
RECODE meaned (0 thru 1=1) (1.0001 thru 2=2) (2.0001
thru 3=3) (3.0001 thru 4=4) (4.0001 thru 5=5) (5.0001 thru 6=6) (6.0001
thru 7=7) .
EXECUTE .
NOTE:
You will need to go into Variable View to label your new education
variable. With this coding, the labels will be as follows:
1 = less_g9
2 = g9_g12
3 = hs
4 = so_col
5 = assoc
6 = bach
7 = grad