Inhoud blog
  • Waarom leerlingen steeds slechter presteren op Nederlandse scholen; en grotendeels ook toepasselijk op Vlaams onderwijs!?
  • Waarom leerlingen steeds slechter presteren op Nederlandse scholen; en grotendeels ook toepasselijk op Vlaams onderwijs!?
  • Inspectie in Engeland kiest ander spoor dan in VlaanderenI Klemtoon op kernopdracht i.p.v. 1001 wollige ROK-criteria!
  • Meer lln met ernstige gedragsproblemen in l.o. -Verraste en verontwaardigde beleidsmakers Crevits (CD&V) & Steve Vandenberghe (So.a) ... wassen handen in onschuld en pakken uit met ingrepen die geen oplossing bieden!
  • Schorsing probleemleerlingen in lager onderwijs: verraste en verontwaardigde beleidsmakers wassen handen in onschuld en pakken uit met niet-effective maatregelen
    Zoeken in blog

    Beoordeel dit blog
      Zeer goed
      Goed
      Voldoende
      Nog wat bijwerken
      Nog veel werk aan
     
    Onderwijskrant Vlaanderen
    Vernieuwen: ja, maar in continuïteit!
    05-12-2017
    Klik hier om een link te hebben waarmee u dit artikel later terug kunt lezen.7 Deadly Sins in Educational Research

    7 Deadly Sins in Educational Research

    Journal ListJ Grad Med Educv.8(4); 2016 OctPMC5060934 J Grad Med Educ. 2016 Oct; 8(4): 483–487.

    7 Deadly Sins in Educational Research

    Katherine Picho, PhD and Anthony R. Artino, Jr, PhD

    Summary

    High-quality educational research is essential for producing generalizable results that can inform medical education. Although questionable research practices can be found in educational research papers, basic steps can prevent these “sins.” After a study has been published it is quite difficult to determine if, when, and how the findings were influenced by questionable research practices; thus, a proactive approach is best. If spurious findings do find their way into the literature, the consequence is a knowledge base rooted in misleading, exaggerated, or entirely false findings. By avoiding the 7 deadly sins described here, medical education researchers will be in better positions to produce high-quality results that advance the field.

    Concerns over the validity of scientific research have grown in recent years, with considerable evidence indicating that most published research findings in the biomedical sciences are false.4 The major flaws that infect research studies—in education as well as biomedical science—often relate to small samples, small effects, and loosely defined and implemented research designs.4

    While many researchers expect that the scientific literature self-corrects over time, this is not always the case. Indeed, considering the “file drawer effect” (unpublished studies with negative outcomes) and the fact that replication remains an underappreciated and relatively uncommon enterprise,5 self-correction of faulty results may be the exception, not the rule. In response to these challenges, this editorial highlights the most common educational research practices, particularly for quantitative studies, that lead researchers to report misleading, exaggerated, or entirely false findings. The intent of this article is to raise awareness and encourage medical education researchers to avoid the “7 deadly sins” in educational research (box).

    Sins Committed Before Research

    Sin #1: The Curse of the Handicapped Literature Review

    Empirical research is the primary means of theory testing and development. It is also essential for testing practical interventions in authentic educational environments. The literature review is central to this process as it identifies existing strengths, weaknesses, and knowledge gaps in a particular field. The literature review informs key aspects of the research process (ie, research questions, design, and methods) and delineates boundaries within which inferences about findings can be discussed. Consequently, sins committed in the literature review process can have profound effects on every aspect of a study and thus negatively influence study quality.6

    Unfortunately, researchers will often conduct partial reviews that are skewed in favor of their hypotheses. Even more common (and worse) is the practice of conducting the literature review after the study has been completed and the results are known. Such practices allow researchers to selectively use articles and revise hypotheses in support of their results. This is a problem because variation due to randomness, which is an expected part of scientific research, yields a fair number of spurious findings.7 Reformulating hypotheses after results are known is not only a backward approach to the scientific method, but it also increases the likelihood of polluting the field of study with false conclusions based on spurious findings. Such practices could explain why some study findings fail to replicate.8

    Sin #2: Inadequate Power

    In quantitative studies, statistical tests help researchers make inferences about the nature and magnitude of the relationships between independent or predictor variables and outcomes. The extent to which conclusions about these inferences are deemed reasonable is sometimes referred to as statistical conclusion validity.9 In the social sciences, many investigations focus on evaluating group differences on certain phenomena. However, there is always the risk that one could falsely find group differences where they do not exist in the population. This is called a type 1 error, or a false positive.9 Type 1 errors can be minimized by increasing the statistical power of a test, which is the probability of finding a statistically significant difference among groups when such a difference actually exists.10 Statistical power values range from 0 (no power) to 1 (extremely high power). Although increasing power to extremely high values (eg, to a power of 1) might seem like a simple solution to drastically reduce the likelihood of obtaining a false positive, this approach has the unintended consequence of increasing the probability of obtaining a false negative, or a type 2 error.9 Therefore, statistical power must walk a fine line between the 2 ends of the spectrum: high enough to detect true group differences without drastically increasing the risk of making a type 2 error. In educational research, the convention for optimum power is typically 0.8.11

    Power is affected by sample size and the number of hypotheses being tested, among other factors. One study found that most studies in the social sciences, including psychology and education,12 were underpowered. In psychology, the average power of studies was 0.35.12 In medical education, it is not uncommon for quantitative studies to be conducted with sample sizes as low as 20, 15, or even 10 participants. Therefore, it is likely that many medical education research studies are insufficiently powered to detect true differences among groups.

    Power is also affected by the magnitude of the expected effect, such as the size of the differences between 2 groups. Hence, in a given study, low power may stem from small samples and small effects or a combination of both.13 In addition to missing a true difference between groups, low power also reduces the likelihood that a statistically significant result represents a true effect rather than a spurious finding.13 Both of these issues weaken the reliability of findings in a given field. The former may lead to prematurely discarding hypotheses that might advance understanding, and the latter, to spurious findings that cannot be replicated.

    A power analysis should be conducted prior to data collection to avoid these negative consequences. Besides increasing sample size, power can be increased by improving experimental design efficiency, such as through the use of equal cell sample sizes; matching participants; measuring covariates a priori; and correcting for covariates in subsequent analyses.

    Sin #3: Ignoring the Importance of Measurement

    Measurement error weakens the relationship between 2 variables and can also strengthen (or weaken) the relationships among 3 or more variables.9 Using measures that have not been tested, or employing those that have poor psychometric properties, only serves to add more “noise” to the results and potentially taints the field with contradictory or implausible findings.14

    Measurement problems can stem from measurement tools (eg, questionnaires) that underrepresent or overrepresent the construct under study. When a measurement tool is too narrow (eg, in the case of single-item measures), then it likely excludes important aspects of the construct and thus fails to capture the true nature of the phenomenon of interest.14 Measurement problems also occur when the outcome variables (eg, test scores, clerkship grades) are too easy or too difficult. Tasks that are extremely easy or difficult lead to ceiling and floor effects, respectively, which weaken correlations and bias results.

    Go to:

    Sins Committed During Research

    Sin #4: Using the Wrong Statistical Tool

    Scholars have written much about the sins related to statistical analyses in research. The most common involve not checking (or reporting) whether the data meet assumptions of the statistical technique being used. Perhaps the most frequently violated assumption is the assumption that observations are independent. Related to this specific violation is the mistake of treating nondependent data as if they were independent (eg, treating data from 20 participants that are measured 3 times as if data are from 60 participants).15

    The violation of such statistical assumptions has the effect of artificially inflating type 1 errors (false positives), which leads to more statistically significant results than warranted. This outcome threatens the validity of inferences that can be made from statistically significant results and can also result in replication failure. To avoid this pitfall, researchers should verify that their data meet the assumptions of the data analytic technique they intend to use. When statistical assumptions are violated, one should take steps to remedy the problem (eg, transforming non-normal data) or use alternate statistical techniques that are robust to these violations (eg, nonparametric statistics for continuous data that do not follow a normal distribution). Moreover, it can be helpful to consult a statistician early in the research process; such a practice is critical to finding the right statistical tool for the job.

    Sin #5: Merciless Torture of Data and Other Questionable Analysis Practices

    Questionable research practices are prevalent in the social sciences, and medical education is not immune to these problems. Although data fabrication constitutes the extreme end of a continuum, there is evidence that other questionable practices are rampant. Examples of such practices include reporting only results that align with one's hypotheses (“cherry picking”), relaxing statistical significant thresholds to fit results, using 1-sided t tests but failing to mention this in the research report, and wrongly rounding P values upward or downward to fit with a hypothesis (eg, reporting P = .04, when the actual P value is .049).16

    Another popular yet questionable practice is fishing, which refers to mining data for statistically significant findings that do not stem from prespecified hypotheses.9 Fishing increases type 1 error rates and artificially inflates statistical significance. Indeed, it would be a sin to restructure an entire study around findings from a fishing expedition, especially since these findings are more likely to be a product of chance than the result of actual differences in the population. Although findings based on fishing expeditions and other questionable practices generally work to the advantage of the researcher (ie, they improve the chances of reaching a statistically significant result and getting published), they ultimately hurt rather than advance knowledge.

    Go to:

    Sins Committed After Research

    Sin #6: Slavery to the P Value

    The most commonly applied and accepted approach to statistical inference in the social sciences is null hypothesis significance testing,17 where a researcher's hypothesis about group differences on a given construct is tested against the null hypothesis: there are no differences.18 Generally, statistical analyses generate a score that reflects mean group differences for a variable, accompanied by test statistics (t ratios, chi-square analyses, etc) and a probability value (P value). P values represent the probability of obtaining the observed group difference or a more extreme result if said difference did not exist in the population from which the data were sampled.19 To determine statistical significance, P values corresponding to .05 (or less than .05) are usually selected as being indicative of a statistically significant group difference.

    Although a useful tool, P values are not very informative. First, a statistically significant result (ie, rejecting the null hypothesis) does not in any way confirm the researcher's hypotheses, although most times it is falsely perceived and interpreted as such.20,21 Second, extremely large sample sizes (eg, in the thousands) will magnify small group differences; the result may be statistically significant yet practically unimportant due to tiny differences. In educational research, large sample sizes are rare but occasionally are seen when large databases are available (eg, specialty board scores). Researchers should focus on supplementing P value statistics with more informative and practical metrics like effect sizes and confidence intervals around effect sizes. Although such metrics have been underreported,22–24 recent efforts are moving research practices in this direction.12 In fact, many journals now require that these metrics be provided in all quantitative research papers.25,26

    Sin #7: Lack of Transparency in Reporting Results and Maintaining Raw Data

    Although author concerns about word count limits or lack of statistical sophistication may cause inadequate reporting, such practices also serve to cover up questionable research practices. For example, authors sometimes include basic information about descriptive statistics (eg, means) but fail to include standard deviations. To advance medical education, it is critical that authors maintain a high level of transparency in reporting results and retain the integrity of their raw data for later analysis by other investigators (eg, data warehousing and data sharing repositories). Correct reporting and transparency of statistical analyses are important because statistical results from articles are used in meta-analyses. Thus, errors of reporting in primary level studies can lead to errors and bias in meta-analytic findings as well. Researchers should strive to provide full information on basic descriptive statistics (sample sizes, means, and standard deviations) and exact P values, regardless of whether or not they are significant. Last but not least, researchers should fully disclose all of their statistical analyses.


    Summary

    High-quality educational research is essential for producing generalizable results that can inform medical education. Although questionable research practices can be found in educational research papers, basic steps can prevent these “sins.” After a study has been published it is quite difficult to determine if, when, and how the findings were influenced by questionable research practices; thus, a proactive approach is best. If spurious findings do find their way into the literature, the consequence is a knowledge base rooted in misleading, exaggerated, or entirely false findings. By avoiding the 7 deadly sins described here, medical education researchers will be in better positions to produce high-quality results that advance the field.



    Geef hier uw reactie door
    Uw naam *
    Uw e-mail *
    URL
    Titel *
    Reactie * Very Happy Smile Sad Surprised Shocked Confused Cool Laughing Mad Razz Embarassed Crying or Very sad Evil or Very Mad Twisted Evil Rolling Eyes Wink Exclamation Question Idea Arrow
      Persoonlijke gegevens onthouden?
    (* = verplicht!)
    Reacties op bericht (0)



    Archief per week
  • 30/04-06/05 2018
  • 23/04-29/04 2018
  • 16/04-22/04 2018
  • 09/04-15/04 2018
  • 02/04-08/04 2018
  • 26/03-01/04 2018
  • 19/03-25/03 2018
  • 12/03-18/03 2018
  • 05/03-11/03 2018
  • 26/02-04/03 2018
  • 19/02-25/02 2018
  • 12/02-18/02 2018
  • 05/02-11/02 2018
  • 29/01-04/02 2018
  • 22/01-28/01 2018
  • 15/01-21/01 2018
  • 08/01-14/01 2018
  • 01/01-07/01 2018
  • 25/12-31/12 2017
  • 18/12-24/12 2017
  • 11/12-17/12 2017
  • 04/12-10/12 2017
  • 27/11-03/12 2017
  • 20/11-26/11 2017
  • 13/11-19/11 2017
  • 06/11-12/11 2017
  • 30/10-05/11 2017
  • 23/10-29/10 2017
  • 16/10-22/10 2017
  • 09/10-15/10 2017
  • 02/10-08/10 2017
  • 25/09-01/10 2017
  • 18/09-24/09 2017
  • 11/09-17/09 2017
  • 04/09-10/09 2017
  • 28/08-03/09 2017
  • 21/08-27/08 2017
  • 14/08-20/08 2017
  • 07/08-13/08 2017
  • 31/07-06/08 2017
  • 24/07-30/07 2017
  • 17/07-23/07 2017
  • 10/07-16/07 2017
  • 03/07-09/07 2017
  • 26/06-02/07 2017
  • 19/06-25/06 2017
  • 05/06-11/06 2017
  • 29/05-04/06 2017
  • 22/05-28/05 2017
  • 15/05-21/05 2017
  • 08/05-14/05 2017
  • 01/05-07/05 2017
  • 24/04-30/04 2017
  • 17/04-23/04 2017
  • 10/04-16/04 2017
  • 03/04-09/04 2017
  • 27/03-02/04 2017
  • 20/03-26/03 2017
  • 13/03-19/03 2017
  • 06/03-12/03 2017
  • 27/02-05/03 2017
  • 20/02-26/02 2017
  • 13/02-19/02 2017
  • 06/02-12/02 2017
  • 30/01-05/02 2017
  • 23/01-29/01 2017
  • 16/01-22/01 2017
  • 09/01-15/01 2017
  • 02/01-08/01 2017
  • 26/12-01/01 2017
  • 19/12-25/12 2016
  • 12/12-18/12 2016
  • 05/12-11/12 2016
  • 28/11-04/12 2016
  • 21/11-27/11 2016
  • 14/11-20/11 2016
  • 07/11-13/11 2016
  • 31/10-06/11 2016
  • 24/10-30/10 2016
  • 17/10-23/10 2016
  • 10/10-16/10 2016
  • 03/10-09/10 2016
  • 26/09-02/10 2016
  • 19/09-25/09 2016
  • 12/09-18/09 2016
  • 05/09-11/09 2016
  • 29/08-04/09 2016
  • 22/08-28/08 2016
  • 15/08-21/08 2016
  • 25/07-31/07 2016
  • 18/07-24/07 2016
  • 11/07-17/07 2016
  • 04/07-10/07 2016
  • 27/06-03/07 2016
  • 20/06-26/06 2016
  • 13/06-19/06 2016
  • 06/06-12/06 2016
  • 30/05-05/06 2016
  • 23/05-29/05 2016
  • 16/05-22/05 2016
  • 09/05-15/05 2016
  • 02/05-08/05 2016
  • 25/04-01/05 2016
  • 18/04-24/04 2016
  • 11/04-17/04 2016
  • 04/04-10/04 2016
  • 28/03-03/04 2016
  • 21/03-27/03 2016
  • 14/03-20/03 2016
  • 07/03-13/03 2016
  • 29/02-06/03 2016
  • 22/02-28/02 2016
  • 15/02-21/02 2016
  • 08/02-14/02 2016
  • 01/02-07/02 2016
  • 25/01-31/01 2016
  • 18/01-24/01 2016
  • 11/01-17/01 2016
  • 04/01-10/01 2016
  • 28/12-03/01 2016
  • 21/12-27/12 2015
  • 14/12-20/12 2015
  • 07/12-13/12 2015
  • 30/11-06/12 2015
  • 23/11-29/11 2015
  • 16/11-22/11 2015
  • 09/11-15/11 2015
  • 02/11-08/11 2015
  • 26/10-01/11 2015
  • 19/10-25/10 2015
  • 12/10-18/10 2015
  • 05/10-11/10 2015
  • 28/09-04/10 2015
  • 21/09-27/09 2015
  • 14/09-20/09 2015
  • 07/09-13/09 2015
  • 31/08-06/09 2015
  • 24/08-30/08 2015
  • 17/08-23/08 2015
  • 10/08-16/08 2015
  • 03/08-09/08 2015
  • 27/07-02/08 2015
  • 20/07-26/07 2015
  • 13/07-19/07 2015
  • 06/07-12/07 2015
  • 29/06-05/07 2015
  • 22/06-28/06 2015
  • 15/06-21/06 2015
  • 08/06-14/06 2015
  • 01/06-07/06 2015
  • 25/05-31/05 2015
  • 18/05-24/05 2015
  • 11/05-17/05 2015
  • 04/05-10/05 2015
  • 27/04-03/05 2015
  • 20/04-26/04 2015
  • 13/04-19/04 2015
  • 06/04-12/04 2015
  • 30/03-05/04 2015
  • 23/03-29/03 2015
  • 16/03-22/03 2015
  • 09/03-15/03 2015
  • 02/03-08/03 2015
  • 23/02-01/03 2015
  • 16/02-22/02 2015
  • 09/02-15/02 2015
  • 02/02-08/02 2015
  • 26/01-01/02 2015
  • 19/01-25/01 2015
  • 12/01-18/01 2015
  • 05/01-11/01 2015
  • 29/12-04/01 2015
  • 22/12-28/12 2014
  • 15/12-21/12 2014
  • 08/12-14/12 2014
  • 01/12-07/12 2014
  • 24/11-30/11 2014
  • 17/11-23/11 2014
  • 10/11-16/11 2014
  • 03/11-09/11 2014
  • 27/10-02/11 2014
  • 20/10-26/10 2014
  • 13/10-19/10 2014
  • 06/10-12/10 2014
  • 29/09-05/10 2014
  • 22/09-28/09 2014
  • 15/09-21/09 2014
  • 08/09-14/09 2014
  • 01/09-07/09 2014
  • 25/08-31/08 2014
  • 18/08-24/08 2014
  • 11/08-17/08 2014
  • 04/08-10/08 2014
  • 28/07-03/08 2014
  • 21/07-27/07 2014
  • 14/07-20/07 2014
  • 07/07-13/07 2014
  • 30/06-06/07 2014
  • 23/06-29/06 2014
  • 16/06-22/06 2014
  • 09/06-15/06 2014
  • 02/06-08/06 2014
  • 26/05-01/06 2014
  • 19/05-25/05 2014
  • 12/05-18/05 2014
  • 05/05-11/05 2014
  • 28/04-04/05 2014
  • 14/04-20/04 2014
  • 07/04-13/04 2014
  • 31/03-06/04 2014
  • 24/03-30/03 2014
  • 17/03-23/03 2014
  • 10/03-16/03 2014
  • 03/03-09/03 2014
  • 24/02-02/03 2014
  • 17/02-23/02 2014
  • 10/02-16/02 2014
  • 03/02-09/02 2014
  • 27/01-02/02 2014
  • 20/01-26/01 2014
  • 13/01-19/01 2014
  • 06/01-12/01 2014
  • 30/12-05/01 2014
  • 23/12-29/12 2013
  • 16/12-22/12 2013
  • 09/12-15/12 2013
  • 02/12-08/12 2013
  • 25/11-01/12 2013
  • 18/11-24/11 2013
  • 11/11-17/11 2013
  • 04/11-10/11 2013
  • 28/10-03/11 2013
  • 21/10-27/10 2013

    E-mail mij

    Druk op onderstaande knop om mij te e-mailen.


    Gastenboek

    Druk op onderstaande knop om een berichtje achter te laten in mijn gastenboek


    Blog als favoriet !

    Klik hier
    om dit blog bij uw favorieten te plaatsen!


    Blog tegen de wet? Klik hier.
    Gratis blog op https://www.bloggen.be - Bloggen.be, eenvoudig, gratis en snel jouw eigen blog!