2. Planning of statistical investigations

2.1. Purpose of the study

• description
• exploration
• explanation
• predection or control

Study question or problem statement:

1. Define a specific problem area.
2. Review the relevant scientific literature.
3. Examine the problem's potential significance to your domain.
4. Pragmatically examine the feasibility of studying the reseach problem.

Good problem statement

1. clearly identifies the variables under consideration,
2. clearly expresses the variables relationships to each other,
3. specifies the nature of the population,
4. implies the possibility of empirical testing.

2.2 Study design

1. True experiments
2. Quasi-experiments
3. Nonexperimental research

2.3 Reliability and validity

Reliability

1. stability
2. internal consistency
3. equivalence
4. other criteria

• efficiescy
• sensitivity
• objectivity
• comprehensibility

Controlling intrinsic factors

1. randomization
2. homogeneity
3. bloking
4. matching
5. covariates

Threats to internal validity

1. history
2. selection
3. maturation
4. testing and instrumentation

2.4 Sampling

Population

Sample

Element / unit

Eligibility criteria

Sample size

Nonprobable vs probable sampling

Nonprobable sampling   Probable sampling  

Convienence            Simple random      
(accidental)                              

Quota                  Stratified         

Purposive              Cluster            
(judgemental)                             

                       Systematic         

2.5 Variables

1. dependent variable (outcome, response)
2. independent variable (predictor, factor, carrier, explanatory variable)
3. extraneous variable (blocking variable)
4. combinied variable (sumvariamle, index, scale)
5. typology

Variables/ samples can be related to each other (siblings, twins) or they are dependent.

2.6 Levels of measurement

1. Nominal
2. Ordinal
3. Interval
4. Ratio

In general, the more detailed you measure, the better!

1. High level of measurement generally yields more information.
2. More powerful and sensitive analytic procedures can be used.
3. When one moves from a higher level to a lower level of measurement there is always an information loss.
4. When one has information at one level, one cal always manipulate the data to arrive at a lower level.

2.7 Statistical procedures

Parametric statistics

• assumptions about the distribution of the variables
• the estimation of the parameter
• the use of at least interval measures

Nonparametric statistics

• less restrictive assumptions
• "distribution-free" statistics
• data measured on a nominal or an ordinal scale
• small sample sizes

Some pages about the nonparametrical tests by the book Siegel, S & Castellan, H.J. (1988). Nonparametric statistics for the behavioral sciences. 2. ed. McGraw Hill, Singapore.