Transcript RUN

EPIB698D Lecture 3
Raul Cruz-Cano
Spring 2013
Proc Univariate
• The UNIVARIATE procedure provides data
summarization on the distribution of numeric variables.
PROC UNIVARIATE <option(s)>;
Var variable-1 variable-n;
Run;
Options:
PLOTS : create low-resolution stem-and-leaf, box, and
normal probability plots
NORMAL: Request tests for normality
data blood;
INFILE 'C:\blood.txt';
INPUT subjectID $ gender $ bloodtype $
age_group $ RBC WBC cholesterol;
run;
proc univariate data =blood ;
var cholesterol;
run;
OUTPUT (1)
The UNIVARIATE Procedure
Variable: cholesterol
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
795
Sum Observations
160141
Variance
2488.6844
Kurtosis
-0.0706044
Corrected SS
1976015.41
Std Error Mean 1.76929947
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
N - This is the number of valid
observations for the variable. The
total number of observations is the
sum of N and the number of missing
values.
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
Moments Moments are
statistical
summaries of a
distribution
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
Sum Weights - A numeric variable can be
specified as a weight variable to weight the
values of the analysis variable. The default
weight variable is defined to be 1 for each
observation. This field is the sum of
observation values for the weight variable
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
Sum Observations - This is the sum of observation values. In case
that a weight variable is specified, this field will be the weighted
sum. The mean for the variable is the sum of observations divided
by the sum of weights.
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
Std Deviation - Standard deviation is the square root of the
variance. It measures the spread of a set of observations. The
larger the standard deviation is, the more spread out the
observations are.
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
Variance - The variance is a measure of variability. It is the sum of
the squared distances of data value from the mean divided by N-1.
We don't generally use variance as an index of spread because it is
in squared units. Instead, we use standard deviation.
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
Skewness - Skewness measures the degree and direction of
asymmetry. A symmetric distribution such as a normal
distribution has a skewness of 0, and a distribution that is
skewed to the left, e.g. when the mean is less than the
median, has a negative skewness.
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
(1)Kurtosis - Kurtosis is a measure of the heaviness of the tails of a
distribution. In SAS, a normal distribution has kurtosis 0.
(2) Extremely nonnormal distributions may have high positive or negative
kurtosis values, while nearly normal distributions will have kurtosis values
close to 0.
(3) Kurtosis is positive if the tails are "heavier" than for a normal distribution
and negative if the tails are "lighter" than for a normal distribution.
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Uncorrected Sum of
Square Distances
from the Mean This is the sum of
squared data
values.
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
Corrected SS - This is the sum of squared
distance of data values from the mean.
This number divided by the number of
observations minus one gives the variance.
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
(1)Coeff Variation - The coefficient of variation is another way
of measuring variability.
(2)It is a unitless measure.
(3)It is defined as the ratio of the standard deviation to the
mean and is generally expressed as a percentage.
(4) It is useful for comparing variation between different
variables.
OUTPUT (1)
Moments
N
Mean
Std Deviation
Skewness
Uncorrected SS
Coeff Variation
795
201.43522
49.8867157
-0.0014449
34234053
24.7656371
Sum Weights
Sum Observations
Variance
Kurtosis
Corrected SS
Std Error Mean
795
160141
2488.6844
-0.0706044
1976015.41
1.76929947
(1)Std Error Mean - This is the estimated standard deviation of
the sample mean.
(2)It is estimated as the standard deviation of the sample
divided by the square root of sample size.
(3)This provides a measure of the variability of the sample
mean.
OUTPUT (2)
Location
Variability
Mean 201.4352 Std Deviation
49.88672
Median 202.0000 Variance
2489
Mode 208.0000 Range
314.00000
Interquartile Range 71.00000
NOTE: The mode displayed is the smallest of 2 modes with a count of 12.
Median - The median is a
measure of central tendency.
It is the middle number when
the values are arranged in
ascending (or descending)
order. It is less sensitive than
the mean to extreme
observations.
Mode - The mode is another measure
of central tendency. It is the value
that occurs most frequently in the
variable.
OUTPUT (3)
Location
Variability
Mean 201.4352 Std Deviation
49.88672
Median 202.0000 Variance
2489
Mode 208.0000 Range
314.00000
Interquartile Range 71.00000
NOTE: The mode displayed is the smallest of 2 modes with a count of 12.
Range - The range is a measure
of the spread of a variable. It is
equal to the difference
between the largest and the
smallest observations.
It is easy to compute and easy
to understand.
Interquartile Range - The interquartile
range is the difference between the
upper (75% Q) and the lower quartiles
(25% Q). It measures the spread of a
data set. It is robust to extreme
observations.
OUTPUT (3)
Tests for Location: Mu0=0
Test
-Statistic- -----p Value------
Student's t
t 113.8503 Pr > |t| <.0001
Sign
M 397.5 Pr >= |M| <.0001
Signed Rank S 158205 Pr >= |S| <.0001
OUTPUT (3)
Student's t
t 113.8503 Pr > |t| <.0001
Sign
M 397.5 Pr >= |M| <.0001
Signed Rank S 158205 Pr >= |S| <.0001
(1) Sign - The sign test is a simple nonparametric procedure to test the null
hypothesis regarding the population median.
(2) It is used when we have a small sample from a nonnormal distribution.
(3)The statistic M is defined to be M=(N+-N-)/2 where N+ is the number of
values that are greater than Mu0 and N- is the number of values that are less
than Mu0. Values equal to Mu0 are discarded.
(4)Under the hypothesis that the population median is equal to Mu0,
the sign test calculates the p-value for M using a binomial distribution.
(5)The interpretation of the p-value is the same as for t-test. In our example the
M-statistic is 398 and the p-value is less than 0.0001. We conclude that the
median of variable is significantly different from zero.
OUTPUT (3)
Student's t
t 113.8503 Pr > |t| <.0001
Sign
M 397.5 Pr >= |M| <.0001
Signed Rank S 158205 Pr >= |S| <.0001
(1) Signed Rank - The signed rank test is also known as the Wilcoxon test. It is
used to test the null hypothesis that the population median equals Mu0.
(2) It assumes that the distribution of the population is symmetric.
(3)The Wilcoxon signed rank test statistic is computed based on the rank sum
and the numbers of observations that are either above or below the median.
(4) The interpretation of the p-value is the same as for the t-test. In our
example, the S-statistic is 158205 and the p-value is less than 0.0001. We
therefore conclude that the median of the variable is significantly different from
zero.
OUTPUT (4)
Quantiles (Definition 5)
Quantile
Estimate
100% Max
99%
95%
90%
75% Q3
50% Median
25% Q1
10%
5%
1%
0% Min
331
318
282
267
236
202
165
138
123
94
17
95% - Ninety-five
percent of all values
of the variable are
equal to or less than
this value.
OUTPUT (5)
Extreme Observations
----Lowest-------Highest--Value Obs
Value Obs
17 829
323 828
36 492
328 203
56 133
328 375
65 841
328 541
69
79
331 191
Missing Values
-----Percent Of----Missing
Missing
Value
Count All Obs
Obs
.
205
20.50
100.00
Extreme
Observations This is a list of
the five lowest
and five highest
values of the
variable
PROC FREQ
• PROC FREQ and PROC MEANS have literally been
part of SAS for over 30 years
• Probably THE most used of the SAS analytical
procedures.
• They provide useful information directly and
indirectly and are easy to use, so people run
them daily without thinking about them.
• These procedures can facilitate the construction
of many useful statistics and data views that are
not readily evident from the documentation.
Proc Freq
• PROC FREQ can be used to count frequencies of both
character and numeric variables
• When you have counts for one variable, it is called one-way
frequencies
• When you have two or more variables, the counts are called
two-way, three-way or so on up to n-way frequencies; or
simply cross-tabulations
• Syntax:
Proc freq ;
Table(s) variable-combinations;
• To produce one-ways frequencies, just put variable name
after “TABLES”; To produced cross-tabulations, put an
asterisk (*) between the variables
22
Proc Freq
• The blood.txt data contain information of 1000 subjects.
The variables include: subject ID, gender, blood_type, age
group, red blood cell count, white blood cell count, and
cholesterol.
• Here is the data with first few subjects:
1 Female AB Young 7710 7.4 258
2 Male AB Old
6560 4.7 .
3 Male A Young 5690 7.53 184
4 Male B Old
6680 6.85 .
5 Male A Young
. 7.72 187
• We want to derive frequencies of gender, age group, and
blood type.
23
PROC FREQ options
• Nocol: Suppress the column percentage for
each cell
• Norow: Suppress the row percentage for each
cell
• Nopercent: Suppress the percentages in
crosstabulation tables, or percentages and
cumulative percentages in one-way frequency
tables and in list format
24
Proc Freq
data blood;
infile "c:\blood.txt";
input ID Gender $ Blood_Type $ Age_Group $
RBC WBC cholesterol;
run;
proc freq data=blood;
tables Gender
Blood_Type;
tables Gender * Age_Group * Blood_Type /
nocol norow nopercent;
run;
25
Continuous Values
• You would produce literally hundreds of categories for each value of
a continuous variable (tuition, math SAT score, salary, age, etc).
• You need to transform the numeric variables into categorical
variables.
proc freq data=blood ;
tables RBC*WBC /norow nocol nopercent ;
run;
data blood;
length rbc_group $ 12 wbc_group $12;
set blood;
if 0 <= RBC <6375 then rbc_group ="Low";
else if 6375 <= RBC < 7040 then rbc_group ="Medium Low";
else if 7040 <= RBC < 7710 then rbc_group ="Medium High";
else rbc_group =" High";
proc univariate data=blood ;
var RBC WBC;
run;
if 0 <= WBC <4.84 then wbc_group ="Low";
else if 4.84 <= WBC < 5.52 then wbc_group ="Medium Low";
else if 5.52 <= WBC < 6.11 then wbc_group ="Medium High";
else wbc_group =" High";
run;
proc freq data=blood ;
tables rbc_group*wbc_group;
run;
PROC FREQ options
• Missprint: Display missing value frequencies
• Missing: Treat missing values as nonmissing
data one;
input A Freq;
datalines;
12
22
.2
;
Run;
proc freq data=one;
tables A;
title 'Default';
run;
proc freq data=one;
tables A / missprint;
title 'MISSPRINT Option';
run;
proc freq data=one;
tables A / missing;
title 'MISSING Option';
run;
27
PROC FREQ options
• The order option orders the values of the
frequency and crosstabulation table variables
according to the specified order, where:
– Data: orders values according to their order in the
input dataset
– Formatted: orders values by their formatted values
– freq: orders values by descending frequency count
– Internal: orders values by their unformatted values
28
PROC FREQ output
Creates an output data set with frequencies, percentages, and expected cell
frequencies
• Out=: Specify an output data set to contain variable values and frequency
counts
data blood;
infile "c:\blood.txt";
input ID Gender $ Blood_Type $ Age_Group $ RBC WBC cholesterol;
run;
proc freq data=blood order = freq;
tables Blood_Type / out = proccount;;
run;
proc print data = proccount;
run;
out goes here not in the proc step
29
PROC FREQ options
• Some of the statistics available through the
FREQ procedure include:
– Chisq provides chi-square tests of independence
of each stratum and computes measures of
association.
– A significant Chi Square (p<.001, for example)
indicates that there is a strong dependence
between the variables
proc freq data=blood ;
tables rbc_group*wbc_group/norow nocol nopercent chisq;
run;
Proc corr
The CORR procedure is a statistical procedure for numeric
random variables that computes correlation statistics (The
default correlation analysis includes descriptive statistics,
Pearson correlation statistics, and probabilities for each
analysis variable).
PROC CORR options;
VAR variables;
WITH variables;
BY variables;
Proc corr data=blood;
var RBC WBC cholesterol;
run;
Proc Corr Output
Simple Statistics
Variable
RBC
WBC
cholesterol
N
908
916
795
Mean
Std Dev
7043
1003
5.48353
0.98412
201.43522 49.88672
Sum
Minimum
6395020
5023
160141
4070
1.71000
17.00000
Maximum
10550
8.75000
331.00000
N - This is the number of valid (i.e., non-missing) cases used in the correlation. By
default, proc corr uses pairwise deletion for missing observations, meaning that a
pair of observations (one from each variable in the pair being correlated) is
included if both values are non-missing. If you use the nomiss option on the proc
corr statement, proc corr uses listwise deletion and omits all observations with
missing data on any of the named variables.
Proc Corr
Pearson Correlation Coefficients
Prob > |r| under H0: Rho=0
Number of Observations
RBC
RBC
1.00000
908
P-value
H0: Corr=0
WBC
cholesterol
Number of
observations
-0.00203
0.9534
833
0.06583
0.0765
725
WBC
cholesterol
-0.00203
0.9534
833
1.00000
916
0.02496
0.5014
728
0.06583
0.0765
725
0.02496
0.5014
728
1.00000
795
Pearson Correlation
Coefficients - measure
the strength and
direction of the linear
relationship between
the two variables.
The correlation
coefficient can range
from -1 to +1, with -1
indicating a perfect
negative correlation, +1
indicating a perfect
positive correlation, and
0 indicating no
correlation at all.
SAS ODS
(Output Delivery System)
• ODS is a powerful tool that can enhance the efficiency of statistical
reporting and meet the needs of the investigator.
• To create output objects that can be send to destinations such as HTML,
PDF, RTF (rich text format), or SAS data sets.
• To eliminate the need for macros that used to convert standard SAS
output to a Microsoft Word, or HTML document
Fish Measurement Data
The data set contains 35 fish from the species Bream caught in
Finland's lake Laengelmavesi with the following measurements:
• Weight (in grams)
• Length3 (length from the nose to the end of its tail, in cm)
• HtPct (max height, as percentage of Length3)
• WidthPct (max width, as percentage of Length3)
ods graphics on;
title 'Fish Measurement Data';
proc corr data=fish1 nomiss plots=matrix(histogram);
var Height Width Length3 Weight3;
run;
ods graphics off;
data Fish1 (drop=HtPct WidthPct);
title 'Fish Measurement Data';
input Weight Length3 HtPct WidthPct @@;
Weight3= Weight**(1/3);
Height=HtPct*Length3/100;
Width=WidthPct*Length3/100;
datalines;
242.0 30.0 38.4 13.4 290.0 31.2 40.0 13.8
340.0 31.1 39.8 15.1 363.0 33.5 38.0 13.3
430.0 34.0 36.6 15.1 450.0 34.7 39.2 14.2
500.0 34.5 41.1 15.3 390.0 35.0 36.2 13.4
450.0 35.1 39.9 13.8 500.0 36.2 39.3 13.7
475.0 36.2 39.4 14.1 500.0 36.2 39.7 13.3
500.0 36.4 37.8 12.0
. 37.3 37.3 13.6
600.0 37.2 40.2 13.9 600.0 37.2 41.5 15.0
700.0 38.3 38.8 13.8 700.0 38.5 38.8 13.5
610.0 38.6 40.5 13.3 650.0 38.7 37.4 14.8
575.0 39.5 38.3 14.1 685.0 39.2 40.8 13.7
620.0 39.7 39.1 13.3 680.0 40.6 38.1 15.1
700.0 40.5 40.1 13.8 725.0 40.9 40.0 14.8
720.0 40.6 40.3 15.0 714.0 41.5 39.8 14.1
850.0 41.6 40.6 14.9 1000.0 42.6 44.5 15.5
920.0 44.1 40.9 14.3 955.0 44.0 41.1 14.3
925.0 45.3 41.4 14.9 975.0 45.9 40.6 14.7
950.0 46.5 37.9 13.7
;
run;
Weighted Data
Problem:
Pct. of Voting
Population
Minority
Voters
White
Voters
Pct. of People who have a
phone
Minority
Voters
White
Voters
Solution: Give more “weight” to the minority people with telephone
Weighted Data
Not limited to 2 categories
Pct. of Voting
Population
Pct. of People who have a
phone
Minority/Dem.
Minority/Dem.
Minority/Rep.
Minority/Rep.
White /Dem
White /Dem
White /Rep
White /Rep
How many categories? As many as there are significant
Proportion
Suppose minority voters are 1/3 of the voting population but only 1/6 of the people with
phone
1
1
?   ?  2
6
3
5
2
4
 ?   ?   .8
6
3
5
Needless to say that in reality this is a much more complex issue
Which weight we need to use?
•
Oversimplified example (don’t take seriously)
Pct. of People who have a
phone
Minority
Voters
White
Voters
Pct. of Voting Population in 2008
Minority
Voters
O
White
Voters
Pct. of Voting Population in 2010
Minority Voters
White
Voters
M
Proportion
Suppose minority voters are 1/3 of the voting population but only 1/6 of the people with
phone
1.
2.
3.
4.
100 minority + 500 white answer the phone survey
75 Minority will vote for candidate X
250 White will votes for candidate X
Non-Weighted Conclusion: 325/600 =54.16% of the voters will vote for
candidate X
5. Weighted Conclusion:
a) 75 minority = 75% of minority with phone=>(.75)*(1/6)=12.5% of people
with phone * 2 weight= 25% pct of voting population
b) 250 white = 50% of white people with phone =>(.5)*(5/6)= 41.66% of
people with phone * .8 weight =>33.33%
c) 25% +33.33%=58.33%
SAS Weighted Mean
proc means data=sashelp.class;
var height;
run;
proc means data=sashelp.class;
weight weight;
var height;
run;
Both PRC FREQ and PROC CORR
also allow to use weights using
the statement weight.
Another (better?) approach for weighted data
•
Experimental design data have all the properties that we learned about in
statistics classes.
– The data are going to be independent
– Identically-distributed observations with some known error distribution
– There is an underlying assumption that the data come to use as a finite
number of observations from a conceptually infinite population
– Simple random sampling without replacement for the sample data
•
Sample survey data,
– The sample survey data do not have independent errors.
– The sample survey data may cover many small sub-populations, so we do
not expect that the errors are identically distributed.
– The sample survey data do not come from a conceptually infinite population.
PROC MEANS vs
PROC SURVEYMEANS
1. We have a target population of 647 receipt amounts, classified by the company region. If we need to perform a full audit, and it is too
expensive to perform the full audit on every one of the receipts in the company database, then we need to take a sample.
2. We want to sample the larger receipt amounts more frequently.
3. Sample 'proportional to size'. That means that we choose a multiplier variable, and we make our choices based on the size of that
multiplier. If the multiplier for receipt A is five times the multiplier for receipt B, then receipt A will be five times more likely to be
selected than receipt B. We will use the receipt amount as the multiplier
data AuditFrame (drop=seed);
seed=18354982;
do i=1 to 600;
if i<101 then region='H';
else if i<201 then region='S';
else if i<401 then region='R';
else region='G';
Amount = round ( 9990*ranuni(seed)+10, 0.01);
output;
end;
do i=601 to 617;
if i<603 then region='H';
else if i<606 then region='S';
else if i<612 then region='R';
else region='G';
Amount = round ( 10000*ranuni(seed)+10000, 0.01);
output;
end;
do i = 618 to 647;
If i<628 then region='H';
else if i<638 then region='S';
else if i<642 then region='R';
else region='G';
Amount = round ( 9*ranuni(seed)+1, 0.01);
output;
end;
run;
PROC MEANS vs
PROC SURVEYMEANS
We can perform this sample selection using PROC SURVEYSELECT:
proc surveyselect data=AuditFrame out=AuditSample3 method=PPS seed=39563462 sampsize=100;
size Amount;
run;
This gives us a weighted random sample of size 100. Now we'll just (artificially) create the data set of the audit
results. We will have a validated receipt amount for each receipt. Ideally, the validated amount would be
exactly equal to the listed amount in every case.
data AuditCheck3;
set AuditSample3;
ValidatedAmt = Amount;
if region='S' and mod(i,3)=0 then ValidatedAmt = round(Amount*(.8+.2*ranuni(1234)),0.01);
if region='H' then do;
if floor(Amount/100)=13 then ValidatedAmt=1037.50;
if floor(Amount/100)=60 then ValidatedAmt=6035.30;
if floor(Amount/100)=85 then ValidatedAmt=8565.97;
if floor(Amount/100)=87 then ValidatedAmt=8872.92;
if floor(Amount/100)=95 then ValidatedAmt=9750.05;
end;
diff = ValidatedAmt - Amount;
run;
PROC MEANS vs
PROC SURVEYMEANS
•
•
•
The WEIGHT statement in PROC MEANS allows a user to give some data points more emphasis.
But that isn't the right way to address the weights we have here. We built a sample using a specific sample design, and we
have sampling weights which have a real, physical meaning. A sampling weight for a given data point is the number of receipts
in the target population which that sample point represents.
The primary difference is the inclusion of the TOTAL= option. PROC SURVEYMEANS allows us to compute a Finite Population
Correction Factor and adjust the error estimates accordingly. This factor adjusts for the fact that we already know the answers
for some percentage of the finite population, and so we really only need to make error estimates for the remainder of that
finite population.
proc means data=AuditCheck3 mean stderr clm;
var ValidatedAmt diff;
run;
proc means data=AuditCheck3 mean stderr clm;
var ValidatedAmt diff;
weight SamplingWeight;
run;
proc surveymeans data=AuditCheck3 mean stderr clm total=647;
var ValidatedAmt diff;
weight SamplingWeight;
run;
Are the point estimates different ? What about the confidence intervals?
Household Component of the Medical
Expenditure Panel Survey (MEPS HC)
• The MEPS HC is a nationally representative survey of the
U.S. civilian noninstitutionalized population.
• It collects medical expenditure data as well as information
on demographic characteristics, access to health care,
health insurance coverage, as well as income and
employment data.
• MEPS is cosponsored by the Agency for Healthcare
Research and Quality (AHRQ) and the National Center for
Health Statistics (NCHS).
• For the comparisons reported here we used the MEPS 2005
Full Year Consolidated Data File (HC-097).
• This is a public use file available for download from the
MEPS web site (http://www.meps.ahrq.gov).
Transforming from SAS transport (SSP)
format to SAS Dataset (SAS7BDAT)
• The MEPS is not a simple random sample, its design includes:
–
–
–
–
Stratification
Clustering
Multiple stages of Selection
Disproportionate sampling.
• The MEPS public use files (such as HC-097) include variables for generating
weighted national estimates and for use of the Taylor method for variance
estimation. These variables are:
– person-level weight (PERWT05F on HC-097)
– stratum (VARSTR on HC-097)
– cluster/psu(VARPSU on HC-097).
LIBNAME PUFLIB 'C:\';
FILENAME IN1 'C:\H97.SSP';
PROC XCOPY IN=IN1 OUT=PUFLIB IMPORT;
RUN;
Needed for even better
estimates of the CI
H97.SASBDAT occupies 408MB
vs. 257MB for H97.SSP vs.
14MB for H97.ZIP
PROC SURVEYFREQ Simple Example
SAS7BDAT
PROC SURVEYFREQ DATA= PUFLIB.H97;
TABLES HISPANX*INSCOV05 / ROW;
WEIGHT PERWT05F;
RUN;