Understanding Sampling, Randomization, and Statistical Techniques in Experimental Design -, Study notes of Statistics

Download Understanding Sampling, Randomization, and Statistical Techniques in Experimental Design - and more Study notes Statistics in PDF only on Docsity!

STA 570

SAMPLING

COMPARATIVE STUDIES

AND

RANDOMIZATION

STATISTICS

• ORIGINAL USAGE

“OF THE STATE”

• SCIENTIFIC USAGE

“TECHNIQUES FOR COLLECTING, ANALYZING,

AND DRAWING CONCLUSIONS FROM DATA”

QUANTITATIVE RESEARCH

• SAMPLE SURVEY- Researcher collects information on which have been RANDOMLY selected from a larger group of units.
• OBSERVATIONAL STUDY- Researcher collects information on units which have been subject to different environments or experiences.
• EXPERIMENT- Researcher controls or manipulates units; i.e., actively intervenes by administering a treatment in order to study its effect on the units.

EXAMPLE (SAMPLING)

• Hoff et. Al. (1988) “Sero prevalence of HIV Among Child-bearing Women,” NEJM.

A sample of 30,708 newborns was selected from hospitals providing obstetrical services in Massachusetts between December 3, 1986 and February 1, 1987. Each was tested for HIV.

• UNIVERSE:

(OR POPULATION)

• SAMPLE:

• UNIT:

• VARIABLE:

EXAMPLE (OBSERVATIONAL STUDY):

• A study was run to determine the effect of protein- calorie malnutrition (PCM) on the hair-shaft diameter in children
• The investigator selected 14 well-nourished children, 10 children who had been diagnosed with mild-moderate PCM, 11 children with severe PCM.
• The hair-shaft diameter (mmx10-2) was measured for each child.

• POPULATIONS?

EXPERIMENT: Randomized Trial

COMPLICATIONS OF THE COX-2 INHIBITORS PARECOXIB AND VALIDECOXIB AFTER CARDIAC SURGERY

Nancy A. Nussmeier, M.D., Andrew A. Whelton, M.D., Mark T. Brown, M.D., Richard M. Langford, F.R.C.A., Andreas Hoeft, M.D., Joel L. Parlow, M.D., Steven W. Boyce, M.D., and Kenneth M. Verburg, Ph.D. NEJM VOLUME 352:1081-1091,March 17, 2005

ABSTRACT Background Valdecoxib and its intravenous prodrug Parecoxib are used to treat postoperative pain but May involve risk after coronary-artery bypass grafting CABG. METHODS in this randomized, double-blind study Involving 10 days of treatment and 30 days of follow-up, 1671patients were randomly assigned to receive 1.) Intravenous parecoxib for at least 3 days, followed By oral valdecoxib through day 10; 2.) intravenous placebo Followed by oral valddecoxib;3.) or placebo for 10 days. All patients had access to standard opoid medications. The Primary end point was the frequency of predefined adverse Events,including cardiovascular events, renal failure or Dysfunction, gastroduodenal ulceration, and wound-healing complications.

EXAMPLE (AZT CONTINUED)

POPULATIONS FOR EXPERIMENTS

Population #

Population of all Aids Patients (Meeting inclusion criterion)-All Patients would receive AZT

Population #

Population of all Aids Patients (meeting inclusion Criterion)-All patients would received Placebo

EXPERIMENTATION VS.OBSERVATION

Observational study- researcher collects information units which have been subject to different environments and/or experiences

Experiment- researcher controls or manipulates

units; i.e., actively intervenes by administering a

treatment in order to supply its effect on units.

For example,

1.) Treatment Æ Observation

2.) observationÆtreatmentÆobservation

Goal of an experiment is to study the effect of

changes in one variable (treatment) on another

variable (measured).

HOW TO SELECT A SIMPLE RANDOM

SAMPLE

• TAG ALL UNITS IN THE AMPLING FRAME OR POPULATION
• MARK OFF NUMBERS IN RANDOM NUMBER TABLE BY NUMBER OF DIGITS REQUIRED

NUMBER OF DIGITS REQUIRED*

***** Always tag starting at zero

Total # of Units # of Digits Required 1-10 1 11-100 2 101-1000 3

. 4 . 5 . 6

SIMPLEST EXPERIMENTAL DESIGN

{PARALLEL GROUPS DESIGN COMPLETELY

RANDOMIZED DESIGN}

• Compare “a” treatments
• Number of experimental units per treatment

n 1 , n 2 ,….,na

EVERY GROUP OF ni UNITS IS EQUALLY LIKELY TO BE ASSIGNED TO THE ith TREATMENT GROUP.

EXAMPLE 1 (LITTLE, 1989)

Nine depressed inpatients have consented to be participants in a randomized trial to assess the psychiatric response to MPH (40 mg orally). Five patients are to be allocated to MPH and four to placebo.

PATIENTS* MATTHEW SMITH CARR NEECE CROWLEY SANDS GODDARD GREENE LIST

QUESTION: Which patients are to be allocated to MPH? ANSWER: Use random number table to select a sample of size five from the nine.

*Names Fictitious

Example: Randomly allocate 5 units to each of three

Treatment groups, say A, B, and C.

• Use random number table
• First five selected receive treatment A
• Next five selected receive treatment B
• Remaining five selected receive treatment C

RANDOMIZATION WITH MORE THAN

TWO TREATMENT GROUPS

STUDY PROBLEM: RANDOMIZATION

SCHEDULE

Starting at line 03, complete the randomization schedule for the scoliosis experiment as described in Example 2.

Order of Entry Treatment Group

1 __ 2 __ 3 __ 4 __ 5 __ 6 __ 7 __ 8 __ 9 __ 10 __ 11 __ 12 __ 13 __ 14 __ 15 __

WHY RANDOMIZE?

1. Create “equivalent” groups
2. Avoid hidden biases and confounding

Confounding- effects of two (or more) factors on a response variable are confounded when the effects cannot be distinguished one from another

Bias- effect of a factor is consistently overestimated (or underestimated)

3. Validity of statistical tests