Training on survey data analysis
Indian Institute of Psychometry, Kolkata invited me to conduct summer training program on survey data analysis from 11-12, 14-17th May, 2009.
I have learned many things through this training. Some of them are given below:
Teaching aids:
1. One computer for each participant;
2. Each computer should be linked to the mother computer through which projection on the screen will be made;
3. Suitable statistical softwares should be in each computer;
4. Room should be air conditioned;
5. Instead of screen, it is better to use big white board on which instructor or trainee can write with marker pen and can clean with specific duster.
Instructor
Subject specific instructor should be there so that extracted results can be properly interpreted.
Problems
It is better to concentrate on data analysis problems instead of theory development. The title should be on data analysis.
For example:
Inter-rater agreement;
Discriminant function analysis;
Predicting group differences;
Assessing Group differences based on correlated variables;
In this training, I have given special emphasis on analysis of data quality.
Data quality is the data sufficiency to satisfy assumptions of statistical models. For example, before use of robust parametric statistics, it is useful to understand whether existing data follows normality assumption or not.
Another one is outlier. Presence of extreme data sometimes causes errors in analysis. So outlier identification is important. In correlation, outlier can be identified through scatter plot analysis.
Box-plot is another tool to understand outliers.
The five figure summary is, therefore:
L = 30 Q1 = 52.75 Q2 = 77.5 Q3 = 101.25 H = 125
Another one:
Multiple box-plot
I have learned many things through this training. Some of them are given below:
Teaching aids:
1. One computer for each participant;
2. Each computer should be linked to the mother computer through which projection on the screen will be made;
3. Suitable statistical softwares should be in each computer;
4. Room should be air conditioned;
5. Instead of screen, it is better to use big white board on which instructor or trainee can write with marker pen and can clean with specific duster.
Instructor
Subject specific instructor should be there so that extracted results can be properly interpreted.
Problems
It is better to concentrate on data analysis problems instead of theory development. The title should be on data analysis.
For example:
Inter-rater agreement;
Discriminant function analysis;
Predicting group differences;
Assessing Group differences based on correlated variables;
In this training, I have given special emphasis on analysis of data quality.
Data quality is the data sufficiency to satisfy assumptions of statistical models. For example, before use of robust parametric statistics, it is useful to understand whether existing data follows normality assumption or not.
Another one is outlier. Presence of extreme data sometimes causes errors in analysis. So outlier identification is important. In correlation, outlier can be identified through scatter plot analysis.
Box-plot is another tool to understand outliers.
The five figure summary is, therefore:
L = 30 Q1 = 52.75 Q2 = 77.5 Q3 = 101.25 H = 125
Another one:
Multiple box-plot
STATISTICAL TOOLS USED BY PARTICIPANTS:
1. Fleiss Kappa for inter rater agreement analysis;
2. Linear Regression;
3. Stepwise multiple regression;
4. Standardized partial regression;
5. Principal component analysis;
6. Analysis of variance;
7. Discriminant function analysis;
Comments