Методология научного творчества = Methodogy of scientific research. Учебно-методическое пособие
.pdfSolomon Four-Group Design – Subjects are randomly assignedintooneoffourgroups: twoexperimentalgroupsandtwo controlgroups.Onlytwogroupsarepretested.Onepretestedgroup and one not-pretested group receive the treatment. All four groups will receive the post-test.
Randomized Block Design [32] – This design is used when there are inherent differences between subjects and possible differences in experimental conditions. If there are a large number of experimental groups, the randomized block design may be used to bring some homogeneity to each group. For example, if a researcherwantedtoexaminetheeffectsofthreedifferentkindsof cough medications on children ages 2-16, the research may want to create age groups (blocks) for the children, realizing that the effects of the medication may depend on age. This is a simple method for reducing the variability among treatment groups.
Crossover Design (also known as Repeat Measures Design)[32] – more than one treatment and different orders of the treatment. The groups compared have an equal distribution of characteristics and there is a high level of similarity among subjects that are exposed to different conditions. In this type of design, the subjects serve as their own control groups.
There are three types of experimental errors: systematic errors and random errors and blunders.
Systematic errors have an identifiable cause, produce results that are consistently too high or low and in theory, can be eliminated.
1)Instrumental – When an instrument itself is flawed and provides inaccurate result
2)Observational – incorrect reading of the measurement
3)Environmental – lab surrounding influences the variables
4)Theoretical – mistake is in the procedure, model or equation.
41
Random Errors [33]
1)are caused by unknown or unpredictable changes in the environment
2)lead to random fluctuations in measurement (about onehalf of the measurements will be too high and one-half too low)
3)do not always have sources that are identifiable
4)can often be quantified by statistical analysis Types:
1)observational – due incorrect reading on a scale (lower or higher)
2)Environmental – unpredictable changes in the environmental conditions (for example vibration) Blunders: not frequent, “one-time" but significant errors.
The total measurement erroris assumed toequal to thesum of random and systematic errors.
There are two another significant types of errors: absolute and relative.
Absoluteerrorisameasureofhowfar'off'ameasurement is a true value or an indication of the uncertainty in a measurement [34]. Absolute Error = Actual Value - Measured Value.
Relative Error expresses how much the absolute error differs from the measured object. Relative error is expressed as fraction or percent.
Relative Error = Absolute Error / Known Value [34] When a measurement is repeated several times the
measured values can be seen grouped around some central value. The central value is called the “mean”, and the spread or deviation of the measured values about the mean is called the “standard deviation”.
The standard deviation of the measured values is represented by the symbol σx and is given by the formula [35]:
42
Any experimental measurement should then be reported in the following form: x=x±σ
The method of least squares
(OLS) is the method of estimating quantities from the results of a set of measurementscontainingrandomerrors [36]. The Method of Least Squares is a procedure, requiring just some calculus and linear algebra, to determine what the “best fit” line is to the data [37].
n |
f (x,a,b,c,...) Ui 2 min. |
|
Source [51] |
||
i 1 |
|
|
|
|
Theorem: The Least Squares Model for a set of data (x1, y1), (x2, y2),..., (xn, yn) passes through the point (x a, y a) where x a is the average of the xi's and ya is the average of the yi's [38].
The essence of the method lies in the fact that the criterion of the quality of the solution under consideration is the sum of squares of errors, which tend to be minimized [39].
Fortheapplicationofthismethod,itisrequiredtocarryout asmanymeasurementsofanunknownrandomvariableaspossible (the higher the higher the accuracy of the solution) and some set of expected solutions from which the best is to be selected. In most cases errors occur in both directions: the estimate may be greater or smaller than the measurement. If we add errors with different signs, then they will be mutually compensated, and as a result, the sum will give us an incorrect idea of the quality of the evaluation. [36]. Often, in order for the final score to have the same dimension as the measured values, the square root is extracted from the sum of error squares.
Field of using:
1) Processing of a scientific experiment data. By a set of measurements with the inevitably "noisy" results (inaccuracy of
43
instrumentsandequipment),ananalyticaldependencecanbebuilt. Iftheexperimentalresultsarewithinthepermissibleerror,thenthe obtained dependence can be used as a physical law for a certain range of input parameters.
2)Forecasting in the sphere of business and economics. If there is a set of data describing, for example, the dynamics of demand from the price of raw materials for a certain period, a regression model that predicts demand next year for different variants of price dynamics can be built.
3)Marketing. Prediction of an advertisement effectiveness according to the past advertising campaigns data. In this case, the input parameters can be a set of keywords and phrases, and the target quantity is the number of responses.
4)Engineering and construction. By deviations from the straight or plane of the parts, the structure, depending on the appliedstresses,temperaturefields,etc.,itispossibletopredictthe values of the deviations in fairly wide ranges of parameters.
Correlation study
There are three types of correlations that are identified[40]:
1.Positive correlation:Positivecorrelationbetweentwovariables is when an increase in one variable leads to an increase in the other and a decrease in one leads to a decrease in the other. For example, the amount of flora bio-productivity might correlate positively with the volume of sedimentation.
2.Negative correlation: Negative correlation is when an increase in one variable leads to a decrease in another and vice versa.For example, the volume of bio-productivity might correlate negatively with radial dryness index.
3.No correlation: Two variables are uncorrelated when a change in one doesn't lead to a change in the other and vice versa. For example, among millionaires, happiness is found to be uncorrelatedtomoney.Thismeansanincreaseinmoneydoesn't lead to happiness.
44
Graphs showing a correlation of -1, 0 and +1 [41]
A correlation coefficient usually varies between +1 and -1. A value close to +1 (for example >0.7) indicates a strong positive correlation while a value close to -1 indicates strong negative correlation. A value near zero shows that the variables are uncorrelated. But it is very important to remember that correlation doesn't imply causation and there is no way to determine or prove causation from a correlational study[40].
The correlation coefficient is calculated by the following formula[42]z
n ai M a bi M b
r= i 1
n a b
where ai and bi are sample data; n - sample size (number of data pairs); Ma and Mb are the corresponding mean values, and a
and b are the mean squares, which are determined by the following formulas:
n ai n bi
Мa= i 1n ; Mb= i 1n ;
45
a= |
n |
ai2 |
M a2 ; b = |
n |
bi2 |
M b2 . |
|
i 1 |
|
i 1 |
|
||||
n |
n |
||||||
|
|
|
|||||
An estimation of the accuracy of the determination of the correlation coefficient r is obtained from
mr=1 nr 2 .
From this formula, it is clear that other things being equal, theerrorincalculatingthecorrelationcoefficientalwaysdecreases with increasing sample size. A sample of at least 50 values is considered representative enough.
VALUE OF SCIENTIFIC WORK
The development and implementation of a method for assessing scientific results are one of the main conditions for optimizing the management of science.
Requirements for significance indicators:
1)should reflect something most significant in the content of the scientific product. Only this quality can ensure the universality of its application, i.e., the possibility of comparing all without exception results.
2)this indicator should provide an opportunity for immediate measurement directly of the scientific result itself, and not the consequences of its use.
To measure the results of scientific research, it is necessary to construct a natural order classification based on an essential basis.
46
Classes of scientific results [15]
The whole set of scientific results is divided by V.S. Libenson and his co-authors [43] into five classes, which exceed each other in the level of concentration of knowledge.
On the basis of - "novelty of scientific information" - also built an order classification of 5 steps.
The breakdown into five is determined by the psychological properties of a person.
The type of work (A1, B1 ... D5) is set at the highest achievement, in each column. The overall performance score is obtained from the product of points.
The quantitative characteristics of the gradations are chosen so that they exclude the same evaluation of works of different types and make the scale more sensitive to value than to the number of works. Messages in which scientific information is completely lacking, in which the truest truths or information is duplicated, are estimated at zero points.
|
|
|
|
|
Table 6 |
|
|
Scale of scientific works importance [43] |
|
||
Class |
of |
scientific |
PointsDegree of novelty |
Points |
|
information |
|
|
|
|
|
А. Description of |
|
|
1. The result is obtained, |
|
|
|
|
which was previously |
|
||
individual, elementary |
|
|
|||
|
fixed in the information |
|
|||
facts (things, properties, |
|
|
|||
|
array but was not known to |
1 |
|||
relationships); Exposition |
1 |
||||
of experience, |
|
|
the author. The existing |
|
|
|
|
scientific information is |
|
||
observations, measurement |
|
|
|||
|
generalized and |
|
|||
results |
|
|
|
|
|
|
|
|
systematized. |
|
|
|
|
|
|
|
|
Б. Elementary analysis of |
|
2. Confirmed or |
|
||
the links between the facts |
|
challenged known ideas |
|
||
with the presence of a |
|
that needed verification. A |
10 |
||
hypothesis, simplex |
2 new version of the solution |
||||
forecast, classification, |
|
is found, which does not |
|
||
explanatory version, or |
|
give any advantages over |
|
||
practical, |
|
|
|
the old one. |
|
47
recommendations of a |
|
|
|
|
private nature. |
|
|
|
|
|
|
3. For the first time, a |
|
|
|
|
connection is found (or a |
|
|
|
|
new link is found) between |
|
|
|
|
the known facts. Known in |
|
|
В. Method (algorithm, |
|
principle provisions are |
|
|
program of measures), |
3 |
extended to new objects, |
100 |
|
device, substance (strain, |
|
resulting in an effective |
|
|
producer) |
|
solution. There are more |
|
|
|
|
simple ways to achieve the |
|
|
|
|
same results. A partial |
|
|
|
|
rational modification (with |
|
|
|
|
signs of novelty) |
|
|
|
|
4. New information has |
|
|
Г. Deep problem |
|
been obtained that |
|
|
|
significantly reduced the |
|
||
development: a |
|
|
||
|
uncertainty of the available |
|
||
multidimensional analysis |
|
|
||
|
knowledge (new facts or |
|
||
of relationships, |
|
|
||
|
patterns have been |
|
||
interdependencies between |
|
|
||
4 |
explained for the first time |
1000 |
||
facts with the presence of |
||||
or for the first time, new |
||||
an explanation, scientific |
|
concepts have been |
|
|
systematization with the |
|
|
||
|
introduced, the structure of |
|
||
construction of a heuristic |
|
|
||
|
content has been revealed). |
|
||
model or a complex |
|
|
||
|
A fundamental |
|
||
forecast |
|
|
||
|
improvement has been |
|
||
|
|
|
||
|
|
made |
|
|
|
|
5. Fundamentally new |
|
|
|
|
scientific information has |
|
|
|
|
been obtained; |
|
|
|
|
Fundamentally new facts |
10000 |
|
Д. Law. Theory. |
5 |
and patterns are revealed; |
||
|
|
A new theory has been |
|
|
|
|
developed. A |
|
|
|
|
fundamentally new device, |
|
|
|
|
substance, method |
|
48
Scientometrics and bibliometrics are methodological approaches in which the scientific literature itself becomes the subject of analysis. In a sense, they could be considered a science of science[44].
Each industry has its own performance indicators. For a long time, the significance of this or that scientific achievement had been estimated by the expert evaluation method. A conception of the researcher’s contribution to science was realized by leading experts in the field after years. This approach leads to undeserved underestimation of the work of those scientists who "do not move in the same circle." One of the main reasons for that is the interlinguas barriers (for example for the Russians, radio was invented by Popov, for Europeans – Marconi, and for Americans – Tesla).
If it is not possible to quickly and qualitatively evaluate the new knowledge brought by scientists, one can try to evaluate the form of their presentation, that is, scientific publications. If a scientist does not publish articles, then his research is unlikely to become public knowledge. So, at first glance - the more articles, the greater the contribution to science.
Despite the fact that "publication activity" is more suited to the role of a symptom of scientific graphomania than to a criterion for assessing the productivity of an individual scientist or a whole scientific project, such an assessment is still widely used.
Even an inattentive reader may notice that the number of articlespublisheddoesnotatallmeantheirquality.Alotoffamous scientists for the whole world have published in their entire life no more than a couple of dozen scientific articles, while even in the most poorly educated university there is an author with a publishing list richer.
In most cases, the problem is solved by relying on the fact that interesting articles are cited more often than no one needs. But this criterion is not without a number of shortcomings:
49
-A truly breakthrough work is always somewhat ahead of its time, so the peak of citations can last for a long period (from several years to several decades).
-The author can artificially provoke a flurry of citations if he touches on any piquant topic
-The very fact of quoting does not mean significance. The practice of specifying bibliographic references in scientific works is such that critical references to works fall into the same list. Imagine that the author ofthe articlewritessomething inthespirit: "... proves the complete inconsistency of the approach [1], in contrastto[2]".So,inthiscase,bothworks[1]and[2]willreceive the same plus sign in the offset.
-Self-citation (they are easily weeded out), as well as citations by students (undergraduates, graduate students) and colleagues, are unlikely to have the same weight as outsiders.
A citation index is an ordered list of cited articles each of which is accompanied by a list of citing articles. The citing article is identified by a source citation, the cited article by a reference citation.
Basic Definitions:
1)The impact factor is a numerical indicator of the importance of a scientific journal, its calculation is based on a three-year period.
For example: for a magazine in 2012 I2011 = A / B, where: A-number of citations during 2011 in journals tracked by the
InstituteofScientificInformation,articlespublishedinthisjournal in 2009-2010; B-number of articles, published in this journal in 2009-2010 [45].
2) The Hirsch index is a scientometric indicator, proposed in 2005 by the American physicist Horne Hirsch (San Diego, California). The scientist with index=5 published h articles, each was referred to at least h times.
3) www.elibrary.ru Russian Scientific Citation Index (RINC) is a national information and analytical system that accumulates more than 2 million publications of Russian authors
50
