Thomas Salzberger, PD Dr. Quantitative Research Methods Closing session December 15, 2020 Starting at 15:15 ...

Page created by Tyrone Logan
 
CONTINUE READING
Thomas Salzberger, PD Dr. Quantitative Research Methods Closing session December 15, 2020 Starting at 15:15 ...
Thomas Salzberger, PD Dr.
thomas.salzberger@wu.ac.at

        Quantitative Research Methods
               Closing session
            December 15, 2020
              Starting at 15:15

 http://statmath.wu.ac.at/courses/m1bw/m1bw_en.html
Thomas Salzberger, PD Dr. Quantitative Research Methods Closing session December 15, 2020 Starting at 15:15 ...
Quantitative Studies

                                                                          Research
                                                         STUDY DESIGN
                                                                          problem,
      Philosophy of science

                                                                        Purpose of the
                                            SAMPLING
                                                                            study

                                          MEASUREMENT        DATA
                                                                           Statistics
                                                           ANALYSIS

                               Psychometrics (Statistics; Metrology)
2                             Conceptual theory (Concept of interest)
Thomas Salzberger, PD Dr. Quantitative Research Methods Closing session December 15, 2020 Starting at 15:15 ...
Successful measurement

“The road from scientific law to scientific measurement
can rarely be traveled in the reverse direction."
The Essential Tension: Selected Studies in Scientific Tradition
and Change (1977), p219.                                                                   Thomas Kuhn
                                                                                            (1922-1996)

"Whether you can observe a thing or not depends on
the theory which you use. It is theory which decides
what can be observed.”
Albert Einstein to Werner Heisenberg during his
1926 Berlin lecture
Edward Fullbrook, “To observe or not to observe: Complementary pluralism in physics and
                                                                                                Ludwik Fleck
economics”, real-world economics review, issue no. 62, 15 December 2012, pp. 20-28,       (1896 Lwiw - 1961)
http://www.paecon.net/PAEReview/issue62/Fullbrook62.pdf

A plea for good conceptual theories:
A qualitative theory of the construct to be measured is
to lead the way to successful measurement.
Thomas Salzberger, PD Dr. Quantitative Research Methods Closing session December 15, 2020 Starting at 15:15 ...
Normal science

 'Normal science' means research firmly based
 upon one or more past scientific achievements,
 achievements that some particular scientific
 community acknowledges for a time as supplying    Thomas Kuhn
 the foundation for its further practice.           (1922-1996)

 The Structure of Scientific Revolutions (1962),
 p10.

 Paradigm for scientific research as
 contextualisation of a study

 One study is one piece in a series of studies
Hypothesis testing

 Conditional probabilities
 type-one error α: chance of rejecting the null hypothesis GIVEN
 it is true
 type-two error β: chance of retaining the null hypothesis GIVEN
 it is not true
 p-value: likelihood of the data (e.g. the difference observed or
 more extreme) GIVEN the null hypothesis

 Probability of the theory GIVEN the data/study
 requires a priori probability of the theory
 Bayes
Analysis

In hypothesis testing, when the p-value is smaller (for example 0.03)
than the designated significance level (for example alpha = 0.05), this
implies ...
□ [a] that the researcher retains the null hypothesis
X [b] that the researcher accepts the alternative hypothesis
□ [c] that the null hypothesis is wrong with a probability of 95%
□ [d] that the alternative hypothesis is true with a probability of 97%

   A=theory, B=data
   P(A) a priori 1% -> P (A| B) a posteriori 14%
   a priori 10% -> a posteriori 64%
   a priori 50% -> a posteriori 94%
Covid Mass Testing

Sensitivity (correct positive) 96.7%
Specificity (correct negative) 99.9%
What is the chance of being infected, if I am tested positive?
What is the chance of being not infected, if I am tested negative?
After 2 days of testing Vienna: about 0.2% tested positive
Prevalence: 0.11%, in other words a priori not infected 99.89%

                       Truly neg       Truly pos   Sum
          Tested neg   99.790          0.004       99.794
          Tested pos   0.100           0.106       0.206
                       99.890          0.11        100

Tested neg: 99.996% neg (99.79/99.794), up from 99.89%
Tested pos: 51.6% pos (0.106/0.206), up from 0.11%
Metaanalysis (in medicine)

     Should pregnant women about to give birth too
      early be given corticosteroids to contribute to
      the development of the baby‘s lungs?

8
Should pregnant women about to give birth
too early be given corticosteroids to
contribute to the development of the baby‘s lungs?
confidence interval            line= no effect
                                                 * (significant!)

  positive effect        negative effect

         * RCT: Randomized Controlled Trial
Should pregnant women about to give birth
too early be given corticosteroids to contribute
to the development of the baby‘s lungs?

                                            *

  * RCT: Randomized Controlled Trial
Measurement

     Over-reliance on a single self-report measure of a construct implies
     which threat to construct validity?
     □ [a] Mono-construct bias
     X [b] Mono-method bias
     □ [c] Mono-operation bias
     □ [d] Mono-study bias
      Mono-method bias refers to your measures or observations, not to your
       programs or causes. (...) With only a single version of a self esteem
       measure, you can’t provide much evidence that you’re really measuring
       self esteem. Your critics will suggest that you aren’t measuring self esteem
       – that you’re only measuring part of it, for instance. -> CONTENT
       VALIDITY
      Solution: try to implement multiple measures of key constructs and try to
       demonstrate (perhaps through a pilot or side study) that the measures
       you use behave as you theoretically expect them to. -> CONVERGENT
       VALIDITY; EXTERNAL VALIDITY
11
Measurement

     Over-reliance on a single self-report measure of a construct implies
     which threat to construct validity?
     □ [a] Mono-construct bias
     X [b] Mono-method bias
     □ [c] Mono-operation bias
     □ [d] Mono-study bias

     Related issue: common-method bias
      All constructs measured using the same method (e.g. self-administered
       questionnaire).
      Correlations are possible due to the same method.
      -> Show that some correlations are close to zero. Use multiple/different
       methods.

12
Measurement

     In the context of classical test theory (true score theory), which of
     the following statements about reliability is/are correct?
     Tick everything that applies!

     X [a] Reliability is the ratio of true score variance and
     observed score variance
     X [b] Reliability is the degree to which a measure is
     consistent
     □ [c] Reliability is the degree to which the researcher can rely on the
     observed scores reflecting what the researcher wants to measure
     -> VALIDITY
     □ [d] Reliability expresses the degree to which a measurement
     instrument is easy to administer

13
Scaling And Indexes

     A Guttman scale …

     Tick everything that applies!

     X [a] is a cumulative model -> sum score meaningful
     X [b] is a deterministic model -> no room for error
     X [c] is a unidimensional model -> one dimension
     □ [d] involves judges rating the items on a scale from 1 to 11

     Note: Rasch model is a probabilistic version of the Guttman model

14
Design

     The regression toward the mean (regression threat) ...

     Multiple options may be correct.

     □ [a] is a consequence of an error in the data analysis
     X [b] is a purely statistical phenomenon
     □ [c] implies that each individual can only go up (or down) from the
     pretest to the posttest -> only at the group level!
     □ [d] can be ruled out when a true effect of the program
     (intervention) exists

15
with your

 from

                         to
               to

                    to
from
                              and finally
You can also read