List four different data combining methods

DEPARTMENT OF COMPUTING AND MATHEMATICS

(TSING YI)

HIGHER DIPLOMA IN APPLIED STATISTICS AND COMPUTING

Unit Name: Statistical Computing

Unit Number: CMFD428

Overall Assessment: Continuous Assessment 100%

Assignment Number: 3

This Assignment: 40% (of 100%)

Hand-out Date: 26 Mar, 1999 (Friday)

Hand-in Date: on 27 April, 1999 (Tuesday) during the lecture

Guidance:

Answer ALL the questions.

Total marks are 100.

Marks will also be given for incomplete SAS programs.

Hand in the hardcopy of the programs and results (no graphical output is required). You should TYPE the numerical results together with the program. Do not give SAS output.

Demonstration in the laboratory is required; marks will be deducted if the oral test during the demonstration is unsatisfactory.

PART A (data management) – 15 marks

List FIVE different data combining methods.
Suggest ONE real life example for each combining method and give the corresponding SAS program (you can create your own datasets for the combinations, if necessary).

PART B (marketing research) – 30 marks

There is a dataset with 100 observations with variables OCCUPATIONS, SEX and INCOME on the server q:\user\practice\hdasc2\sc\a3_partb.sd2.

You are required to draw a sample from this dataset of size equal to 20 in the following manner.

i) Draw a simple random sample of size

(a) approximately 20;

(b) exactly 20.

ii) Draw a systematic sample

(a) without a random start and the first observation is drawn;

(b) with a random start.

iii) Draw a stratified random sample with OCCUPATIONS as the strata.

Hints: Count the number of observations in each occupation.

Select the proportionate sample size (approximately) in each occupation (i.e. stratum).

iv) Draw a quota sample with the control element being Male, i.e., taking the first 10 male respondents.

PART C (simulation and time series) – 25 marks

Randomly generate 365 observations (days) which follow a normal distribution with mean equal to 10 and standard deviation equal to 5.
Based on the data in i), use SAS macro to calculate a n day simple moving average where n and the starting date are input parameters.

PART D (ANOVA) – 30 marks

Two groups of children, one group with Attention Deficit Disorder (ADD) and the other (control) group without ADD, were randomly given either the drug PLACEBO or the drug RITALIN. A measure of activity was made on all the children with the results shown in the table below (Higher Numbers indicate more activity).

GROUP	DRUG	ACTIVITY
ADD	PLACEBO	90
ADD	PLACEBO	88
ADD	PLACEBO	95
CONTROL	PLACEBO	60
CONTROL	PLACEBO	62
CONTROL	PLACEBO	66
ADD	RITALIN	72
ADD	RITALIN	70
ADD	RITALIN	64
CONTROL	RITALIN	86
CONTROL	RITALIN	86
CONTROL	RITALIN	82

Perform a two-way ANOVA (GROUP by DRUG) analysis on these data. Use a Duncan multiple-range test to determine any significant main effects. Draw an interaction plot like the following one (Use OPTIONS PS=25 LS=80;).

Refer to the hardcopy as the following diagram cannot be fixed in html format.

Plot of ACTIVITY*DRUG. Symbol is value of GROUP.

ACTIVITY ‚

100 ˆ

‚ A

‚

‚ A

‚ A C

‚ C

80 ˆ

‚

‚ A

‚ C

‚ C A

60 ˆ C

Šƒƒˆƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒˆƒƒ

PLACEBO RITALIN

DRUG

NOTE: 1 obs hidden.

Write the SAS statements to conduct a t-test between drugs for the ADD and control children respectively.

An alternative approach to i) is to run a one-way ANOVA with each combination of GROUP and DRUG as levels of the independent variable. That is, the four levels would be ADD-PLACEBO, ADD-RITALIN, CONTROL-PLACEBO, and CONTROL-RITALIN. Write a DATA step that creates a new variable (call it LEVEL) which define each of these four groups. Then perform a one-way ANOVA analysis using this new variable as the independent (CLASS) variable. Include the Duncan multiple-range test in your analysis as well.