You are on page 1of 23

What is a Survey ?

A Survey collects information Is a Census from all the population Is a Poll if for political information Is a Sample Survey if from just a sample of a population

Conditions for a Survey


Answer the Objectives Be Unbiased, Accurate Be Generalizable Be Ethical Be Economical

Before a Survey
Define the Questions to be answer. Define the Sampling strategy. Design and Test the Questionnaire. Train the field workers. Define the technique for crossvalidation. Define the final Analysis.

During the Survey


- Verify and crossvalidate the Questionnaire - Check Time table and Budget

After the Survey


- Crosscheck again all the data - Perform the main analysis - Perform any exploratory data analysis.

The Question
The first task is to clearly and concisely define the Main Question of interest as well as the Target Population of the study.

Sampling Unit
The individual A family A group, i.e. a class.

Sampling Frame
Existing listing New listing

Define the Precision or Tolerated error of the estimates : 1% , 5%. Are there any Nuisance factor to be accounted for, which ones are important?

Stratification
Sample size proportional to strata size Equal sample sizes from all strata Sample size inversely proportional to strata variability.

Multiple Steps Sampling

Sampling Units

2nd Units Big Units

Questionnaire
Short and Concise Open Questions Multiple choice Yes/No

Internal Validity
How to check if the subject answers truthfully?
- Ask the same question twice? Marital status Spouse age - Check frequency versus known data

External validity
Are the appropriate questions asked? Are the questions understandable by all the subjects in the sample.

Study of Trends
Cohort Panel same same same new same

new

Interviewers in Follow Up Studies


Preferably be the same. If not the training must be the same. They must be comparable.

Preparing for Validation


Lists of valid responses for each question. Define code for: Missing values. Not applicable. Do not know. Automatically set the Not applicable based on some previous question.

Definition of the Analysis


Specific Objective 1 - Test to answer specific objective 1. Specific Objective 2 - Test to answer specific objective 2. Secondary objectives - Test to investigate secondary

During the Survey


Collect questionnaires daily. Check yesterday questionnaires for missing or invalid answers. Re-interview invalid questionnaires

The Time Table The often forgotten tasks.


Is the number of answered questionnaires according to the predicted for this date?. Is more time needed to fill up the required number of questionnaires? Are more interviewers needed to complete the task in time?

The Interim View


Crosstabulate the descriptor variables. i.e. age, gender, occupation, etc. Are they similar to the known ( or assumed) distribution in the population? Are we getting a representative or a biased sample?. Do we need more advertising of the study in some sectors of the population?

After the Survey


Check for outliers in all variables, both singly and in logical pairs. Are the distributions and scattergrams plausible ? Decide whether to impute or to eliminate the clear mistakes.

Statistical Analysis
Main Analysis Description of the sample Predefined Main Statistical analysis Statistical significance of results Exploratory Analysis Describe test / procedure used Report results

The Report
List of Objectives. Objectives achieved. Sample size estimated. Response rate in the sample. Main Statistical Analysis. Results with significance or Confidence Interval of estimators. Statistical Power of the tests performed. Results of the Exploratory Data Analysis

Conclusions
What was achieved. Did the main Analysis agree with predictions and if not why? Interpretation of the results of the main Analysis. Were any new hypothesis generated by the EDA. Interpretation of the findings by the EDA. Ideas for new research.

You might also like