Statistics Terms

Absolute Value

The absolute value of a number is its distance from zero on the number line. For example, -7 is 7 units away from zero, so its absolute value would be 7. And 7 is also 7 units away from zero, so its absolute value would also be 7.

Thus, the absolute value of a number refers to the magnitude of the number, without regard to its sign. The absolute value of -1 and 1 is 1, the absolute value of -2 and 2 is 2, the absolute value of -3 and 3 is 3, and so on.

Accuracy

Accuracy refers to how close a sample statistic is to a population parameter . Thus, if you know that a sample mean is 99 and the true population mean is 100, you can make a statement about the sample accuracy. For example, you might say the sample mean is accurate to within 1 unit.

Alpha

With respect to estimation problems , alpha refers to the likelihood that the true population parameter lies outside the confidence interval . Alpha is usually expressed as a proportion. Thus, if theconfidence level is 95%, then alpha would equal 1 - 0.95 or 0.05.

With respect to hypothesis tests , alpha refers to significance level , the probability of making a Type I error .

Confidence Interval

Statisticians use a confidence interval to express the degree of uncertainty associated with a samplestatistic. A confidence interval is an interval estimate combined with a probability statement.

For example, suppose a statistician conducted a survey and computed an interval estimate, based on survey data. The statistician might use a confidence level to describe uncertainty associated with the interval estimate. He/she might describe the interval estimate as a "95% confidence interval". This means that if we used the same sampling method to select different samples and computed an interval estimate for each sample, we would expect the true population parameter to fall within the interval estimates 95% of the time.

Confidence intervals are preferred to point estimates and to interval estimates, because only confidence intervals indicate (a) the precision of the estimate and (b) the uncertainty of the estimate.

Estimation

In statistics, estimation refers to the process by which one makes inferences about a population, based on information obtained from a sample. Often, we use sample statistics (e.g., mean, proportion) to estimate population parameters (e.g., mean, proportion).

Alternative Hypothesis

There are two types of statistical hypotheses.

• Null hypothesis. The null hypothesis, denoted by H₀, is usually the hypothesis that sample observations result purely from chance.
  
• Alternative hypothesis. The alternative hypothesis, denoted by H₁ or H_a, is the hypothesis that sample observations are influenced by some non-random cause.

For example, suppose we wanted to determine whether a coin was fair and balanced. A null hypothesis might be that half the flips would result in Heads and half, in Tails. The alternative hypothesis might be that the number of Heads and Tails would be very different. Symbolically, these hypotheses would be expressed as

H₀: p = 0.5
H_a: p <> 0.5

Suppose we flipped the coin 50 times, resulting in 40 Heads and 10 Tails. Given this result, we would be inclined to reject the null hypothesis. That is, we would conclude that the coin was probably not fair and balanced.

Alternative Hypothesis

There are two types of statistical hypotheses.

• Null hypothesis. The null hypothesis, denoted by H₀, is usually the hypothesis that sample observations result purely from chance.
  
• Alternative hypothesis. The alternative hypothesis, denoted by H₁ or H_a, is the hypothesis that sample observations are influenced by some non-random cause.

For example, suppose we wanted to determine whether a coin was fair and balanced. A null hypothesis might be that half the flips would result in Heads and half, in Tails. The alternative hypothesis might be that the number of Heads and Tails would be very different. Symbolically, these hypotheses would be expressed as

H₀: p = 0.5
H_a: p <> 0.5

Suppose we flipped the coin 50 times, resulting in 40 Heads and 10 Tails. Given this result, we would be inclined to reject the null hypothesis. That is, we would conclude that the coin was probably not fair and balanced.

Bias

Bias refers to the tendency of a measurement process to over- or under-estimate the value of a population parameter. In survey sampling, for example, bias would be the tendency of a samplestatistic to systematically over- or under-estimate a population parameter.

Biased Estimate

When the mean of the sampling distribution of a statistic is not equal to a population parameter, that statistic is said to be a biased estimate of the parameter.

Bimodal Distribution

Distributions of data can have few or many peaks. Distributions with one clear peak are calledunimodal, and distributions with two clear peaks are called bimodal, as illustrated in the figures below.

1 2 3 4 5 6 7		1 2 3 4 5 6 7
Unimodal		Bimodal

Confidence Interval

Statisticians use a confidence interval to express the degree of uncertainty associated with a samplestatistic. A confidence interval is an interval estimate combined with a probability statement.

For example, suppose a statistician conducted a survey and computed an interval estimate, based on survey data. The statistician might use a confidence level to describe uncertainty associated with the interval estimate. He/she might describe the interval estimate as a "95% confidence interval". This means that if we used the same sampling method to select different samples and computed an interval estimate for each sample, we would expect the true population parameter to fall within the interval estimates 95% of the time.

Confidence intervals are preferred to point estimates and to interval estimates, because only confidence intervals indicate (a) the precision of the estimate and (b) the uncertainty of the estimate.

Degrees of Freedom

The number of degrees of freedom generally refers to the number of independent observations in a sample minus the number of population parameters that must be estimated from sample data.

For example, the exact shape of a t distribution is determined by its degrees of freedom. When the t distribution is used to compute a confidence interval for a mean score, one population parameter (the mean) is estimated from sample data. Therefore, the number of degrees of freedom is equal to the sample size minus one.

Normal Distribution

The normal distribution is a probability distribution that associates the normal random variable X with a cumulative probability . The normal distribution is defined by the following equation:

Normal equation. The value of the random variable Y is:

Y = [ 1/σ * sqrt(2π) ] * e ^{-(x - μ)2/2σ2}

where X is a normal random variable, μ is the mean, σ is the standard deviation, π is approximately 3.14159, and e is approximately 2.71828.

The graph of the normal distribution depends on two factors - the mean and the standard deviation. The mean of the distribution determines the location of the center of the graph, and the standard deviation determines the height of the graph. When the standard deviation is large, the curve is short and wide; when the standard deviation is small, the curve is tall and narrow. All normal distributions look like a symmetric, bell-shaped curve, as shown below.

The curve on the top is shorter and wider than the curve on the bottom, because the curve on the top has a bigger standard deviation.

Range

The range is a simple measure of variation in a set of random variables. It is difference between the biggest and smallest random variable.

Range = Maximum value - Minimum value

Therefore, the range of the four random variables (3, 5, 5, 7} would be 7 - 3 or 4.

t-Test

A t-test is any hypothesis test in which the test statistic follows Student's t distribution if the null hypothesis is true. Some common t-tests are:

• One-sample t-test. Used to determine whether a hypothesized population mean differs significantly from an observed sample mean. See one-sample t-test example.
  
• Two-sample t-test. Used to determine whether the difference between samples means differs significantly from the hypothesized difference between population means. See two-sample t-test example.
  
• Matched pairs t-test. Used to test the significance of the difference between paired means. SeeMatched pairs t-test example.
  
• Linear regression t-test. Used in simple linear regression to determine whether the slope of regression line differs significantly from zero. See linear regression t-test example.

Variance

The variance is a numerical value used to indicate how widely individuals in a group vary. If individual observations vary greatly from the group mean, the variance is big; and vice versa.

It is important to distinguish between the variance of a population and the variance of a sample. They have different notation, and they are computed differently. The variance of a population is denoted by σ²; and the variance of a sample, by s².

The variance of a population is defined by the following formula:

σ² = Σ ( X_i - X )² / N

where σ² is the population variance, X is the population mean, X_i is the ith element from the population, and N is the number of elements in the population.

The variance of a sample is defined by slightly different formula:

s² = Σ ( x_i - x )² / ( n - 1 )

where s² is the sample variance, x is the sample mean, x_i is the ith element from the sample, and n is the number of elements in the sample. Using this formula, the variance of the sample is an unbiased estimate of the variance of the population.

And finally, the variance is equal to the square of the standard deviation.

Reference

http://stattrek.com/statistics/dictionary.aspx?definition=Degrees_of_freedom

Basic Information on the t-Test

Hypothesis: The hypothesis is a tentative explanation based on observations you have made.  Your observations may have been followed up with a search of the literature for more information before you develop your hypothesis.

Example: Men’s hands are larger than women’s hands OR adding fertilizer to a plant makes it grow better.

Null hypothesis:  The actual null hypothesis is a more formal statement of your original hypothesis.  The null hypothesis is usually written in the following form:  There is no significant difference between population A and population B. 

Example:  There is no significant difference in hand size between males and females.  OR  There is no significant difference in the growth of fertilized plants vs. unfertilized plants.

The reason we write it in this form is that scientists are basically skeptics and their goal is to prove a hypothesis false.  In fact, you can never really prove that a hypothesis is true.  In addition, the null hypothesis is used because it allows you to relate your calculations of the difference between the sample means to a standard of zero.  

The t-Test:  We use this statistical test to compare our sample populations and determine if there is a significant difference between their means. The result of the t-test is a ‘t’ value; this value is then used to determine the p-value (see below).

 If we cannot use a statistical test (doesn’t have to be a t-test) to determine whether a significant difference exists, then it becomes difficult to convince other scientists that your research is worth anything.

P-value: The p-value is the probability that ‘t’ falls into a certain range.  In other words this is the value you use to determine if the difference between the means in your sample populations is significant.  For our purposes, a p-value < 0.05 suggests a significant difference between the means of our sample population and we would reject our null hypothesis.  A p-value > 0.05 suggests no significant difference between the means of our sample populations and we would not reject our null hypothesis.

Types of t-tests:  There are two types of t-tests, the unpaired and paired t-test that we will use in this course.

            Unpaired t-test:  This type of t-test is used when you have independent samples.          In other words your samples are not directly related to one another.  Ex.: Index finger length between males and females.

            Paired t-test:  In this t-test your samples are related.  You collected data before and after some manipulation of your subjects.  Ex.: Pulse before and after 3 cups of coffee.

Reference

http://www.nku.edu/~intsci/sci110/worksheets/basic_ttest_info.html

The T-Test

The t-test assesses whether the means of two groups are statistically different from each other. This analysis is appropriate whenever you want to compare the means of two groups, and especially appropriate as the analysis for the posttest-only two-group randomized experimental design.

Figure 1. Idealized distributions for treated and comparison group posttest values.

Figure 1 shows the distributions for the treated (blue) and control (green) groups in a study. Actually, the figure shows the idealized distribution -- the actual distribution would usually be depicted with a histogram or bar graph. The figure indicates where the control and treatment group means are located. The question the t-test addresses is whether the means are statistically different.

What does it mean to say that the averages for two groups are statistically different? Consider the three situations shown in Figure 2. The first thing to notice about the three situations is that the difference between the means is the same in all three. But, you should also notice that the three situations don't look the same -- they tell very different stories. The top example shows a case with moderate variability of scores within each group. The second situation shows the high variability case. the third shows the case with low variability. Clearly, we would conclude that the two groups appear most different or distinct in the bottom or low-variability case. Why? Because there is relatively little overlap between the two bell-shaped curves. In the high variability case, the group difference appears least striking because the two bell-shaped distributions overlap so much.

Figure 2. Three scenarios for differences between means.

This leads us to a very important conclusion: when we are looking at the differences between scores for two groups, we have to judge the difference between their means relative to the spread or variability of their scores. The t-test does just this.

Statistical Analysis of the t-test

The formula for the t-test is a ratio. The top part of the ratio is just the difference between the two means or averages. The bottom part is a measure of the variability or dispersion of the scores. This formula is essentially another example of the signal-to-noise metaphor in research: the difference between the means is the signal that, in this case, we think our program or treatment introduced into the data; the bottom part of the formula is a measure of variability that is essentially noise that may make it harder to see the group difference. Figure 3 shows the formula for the t-test and how the numerator and denominator are related to the distributions.

Figure 3. Formula for the t-test.

The top part of the formula is easy to compute -- just find the difference between the means. The bottom part is called the standard error of the difference. To compute it, we take the variance for each group and divide it by the number of people in that group. We add these two values and then take their square root. The specific formula is given in Figure 4:

Figure 4. Formula for the Standard error of the difference between the means.

Remember, that the variance is simply the square of the standard deviation.

The final formula for the t-test is shown in Figure 5:

Figure 5. Formula for the t-test.

The t-value will be positive if the first mean is larger than the second and negative if it is smaller. Once you compute the t-value you have to look it up in a table of significance to test whether the ratio is large enough to say that the difference between the groups is not likely to have been a chance finding. To test the significance, you need to set a risk level (called the alpha level). In most social research, the "rule of thumb" is to set the alpha level at .05. This means that five times out of a hundred you would find a statistically significant difference between the means even if there was none (i.e., by "chance"). You also need to determine the degrees of freedom (df) for the test. In the t-test, the degrees of freedom is the sum of the persons in both groups minus 2. Given the alpha level, the df, and the t-value, you can look the t-value up in a standard table of significance (available as an appendix in the back of most statistics texts) to determine whether the t-value is large enough to be significant. If it is, you can conclude that the difference between the means for the two groups is different (even given the variability). Fortunately, statistical computer programs routinely print the significance test results and save you the trouble of looking them up in a table.

The t-test, one-way Analysis of Variance (ANOVA) and a form of regression analysis are mathematically equivalent (see the statistical analysis of the posttest-only randomized experimental design) and would yield identical results.

Refrence

http://www.socialresearchmethods.net/kb/stat_t.php

Measures of Spread

What are measures of spread?

---

Measures of spread describe how similar or varied the set of observed values are for a particular variable (data item). Measures of spread include the range,quartiles and the interquartile range, variance and standard deviation.


When can we measure spread?

---

The spread of the values can be measured for quantitative data, as the variables are numeric and can be arranged into a logical order with a low end value and a high end value.


Why do we measure spread?

---

Summarising the dataset can help us understand the data, especially when the dataset is large. As discussed in the Measures of Central Tendency page, the mode,median, and mean summarise the data into a single value that is typical or representative of all the values in the dataset, but this is only part of the 'picture' that summarises a dataset. Measures of spread summarise the data in a way that shows how scattered the values are and how much they differ from the mean value.

For example:  Dataset A Dataset B 4, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 8 1, 2, 3, 4, 5, 6, 6, 7, 8, 9, 10, 11 The mode (most frequent value), median (middle value*) and mean (arithmetic average) of both datasets is 6. (*note, the median of an even numbered data set is calculated by taking the mean of the middle two observations). If we just looked at the measures of central tendency, we may assume that the datasets are the same. However, if we look at the spread of the values in the following graph, we can see that Dataset B is more dispersed than Dataset A. Used together, the measures of central tendency and measures of spread help us to better understand the data

What does each measure of spread tell us?

---

The range is the difference between the smallest value and the largest value in a dataset. 

Calculating the Range Dataset A 4, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 8 The range is 4, the difference between the highest value (8 ) and the lowest value (4). Dataset B 1, 2, 3, 4, 5, 6, 6, 7, 8, 9, 10, 11 The range is 10, the difference between the highest value (11 ) and the lowest value (1). Dataset A 0 1 2 3 4 5 6 7 8 9 10 11 12 13 Dataset B 0 1 2 3 4 5 6 7 8 9 10 11 12 13 On a number line, you can see that the range of values for Dataset B is larger than Dataset A.

Quartiles divide an ordered dataset into four equal parts, and refer to the values of the point between the quarters. A dataset may also be divided into quintiles (five equal parts) or deciles (ten equal parts).

Quartiles

25% of values

Q1

25% of values

Q2

25% of values

Q3

25% of values

The lower quartile (Q1) is the point between the lowest 25% of values and the highest 75% of values. It is also called the 25th percentile. 

The second quartile (Q2) is the middle of the data set. It is also called the 50th percentile, or the median.

The upper quartile (Q3) is the point between the lowest 75% and highest 25% of values. It is also called the 75th percentile.

Calculating Quartiles Dataset A 4 5 5 Q1 5 6 6 Q2 6 6 7 Q3 7 7 8 As the quartile point falls between two values, the mean (average) of those values is the quartile value: Q1 = (5+5) / 2 = 5 Q2 = (6+6) / 2 = 6 Q3 = (7+7) / 2 = 7 Dataset B 1 2 3 Q1 4 5 6 Q2 6 7 8 Q3 9 10 11 As the quartile point falls between two values, the mean (average) of those values is the quartile value: Q1 = (3+4) / 2 = 3.5 Q2 = (6+6) / 2 = 6 Q3 = (8+9) / 2 = 8.5

The interquartile range (IQR) is the difference between the upper (Q3) and lower (Q1) quartiles, and describes the middle 50% of values when ordered from lowest to highest. The IQR is often seen as a better measure of spread than the range as it is not affected by outliers.

Interquartile Range

25% of values

Q1

25% of values

Q2

25% of values

Q3

25% of values

Calculating the Interquartile Range The IQR for Dataset A is = 2 IQR = Q3 - Q1  = 7 - 5 = 2 The IQR for Dataset B is = 5 IQR = Q3 - Q1  = 8.5 - 3.5 = 5

The variance and the standard deviation are measures of the spread of the data around the mean. They summarise how close each observed data value is to the mean value.

In datasets with a small spread all values are very close to the mean, resulting in a small variance and standard deviation. Where a dataset is more dispersed, values are spread further away from the mean, leading to a larger variance and standard deviation.

The smaller the variance and standard deviation, the more the mean value is indicative of the whole dataset. Therefore, if all values of a dataset are the same, the standard deviation and variance are zero.

The standard deviation of a normal distribution enables us to calculate confidence intervals. In a normal distribution, about 68% of the values are within one standard deviation either side of the mean and about 95% of the scores are within two standard deviations of the mean.

The population Variance σ² (pronounced sigma squared) of a discrete set of numbers is expressed by the following formula:

where:
X_i represents the ith unit, starting from the first observation to the last
μ represents the population mean
N represents the number of units in the population

The Variance of a sample s² (pronounced s squared) is expressed by a slightly different formula:

where:
x_i represents the ith unit, starting from the first observation to the last
x̅ represents the sample mean
n represents the number of units in the sample

The standard deviation is the square root of the variance. The standard deviation for a population is represented by σ, and the standard deviation for a sample is represented by s.

Calculating the Population Variance σ2 and Standard Deviation σ

Dataset A Calculate the population mean (μ) of Dataset A. (4 + 5 + 5 + 5 + 6 + 6 + 6 + 6 + 7 + 7 + 7 + 8) / 12 mean (μ) = 6 Calculate the deviation of the individual values from the mean by subtracting the mean from each value in the dataset  = -2, -1, -1, -1, 0, 0, 0, 0, 1, 1, 1, 2 Square each individual deviation value  = 4, 1, 1, 1, 0, 0, 0, 0, 1,1,1, 4 Calculate the mean of the squared deviation values  = (4 + 1 +1 +1 + 0 + 0 + 0 + 0 +1 +1 +1 + 4) / 12 Variance σ2= 1.17 Calculate the square root of the variance Standard deviation σ = 1.08

Dataset B Calculate the population mean (μ) of Dataset B. (1 + 2 + 3 + 4 + 5 + 6 + 6 + 7 + 8 + 9 + 10 + 11) / 12 mean (μ) = 6 Calculate the deviation of the individual values from the mean by subtracting the mean from each value in the dataset  = -5, -4, -3, -2, -1, 0, 0, 1, 2, 3, 4, 5, Square each individual deviation value  = 25, 16, 9, 4, 1, 0, 0, 1, 4, 9, 16, 25 Calculate the mean of the squared deviation values  = (25 + 16 + 9 + 4 + 1 + 0 + 0 + 1 + 4 + 9 + 16 + 25) / 12 Variance σ2 = 9.17 Calculate the square root of the variance Standard deviation σ = 3.03

The larger Variance and Standard Deviation in Dataset B further demonstrates that Dataset B is more dispersed than Dataset A.

Reference

http://www.abs.gov.au/websitedbs/a3121120.nsf/home/statistical+language+-+measures+of+spread

Scientific Misconduct

Scientific misconduct is the violation of the standard codes of scholarly conduct and ethical behavior in professionalscientific research.

 definitions:(reproduced in The COPE report 1999.)

• Danish definition: "Intention or gross negligence leading to fabrication of the scientific message or a false credit or emphasis given to a scientist"

• Swedish definition: "Intention[al] distortion of the research process by fabrication of data, text, hypothesis, or methods from another researcher's manuscript form or publication; or distortion of the research process in other ways."

Motivation to commit scientific misconduct

Career pressure

Science is still a very strongly career-driven discipline. Scientists depend on a good reputation to receive ongoing support and funding, and a good reputation relies largely on the publication of high-profile scientific papers. Hence, there is a strong imperative to "publish or perish". Clearly, this may motivate desperate (or fame-hungry) scientists to fabricate results.

Ease of fabrication

In many scientific fields, results are often difficult to reproduce accurately, being obscured by noise, artifacts, and other extraneous data. That means that even if a scientist does falsify data, they can expect to get away with it – or at least claim innocence if their results conflict with others in the same field. There are no "scientific police" who are trained to fight scientific crimes; all investigations are made by experts in science but amateurs in dealing with criminals. It is relatively easy to cheat although difficult to know exactly how many scientists fabricate data.

Types of misconducts 

• Fabrication is making up results and recording or reporting them. This is sometimes referred to as "drylabbing".A more minor form of fabrication is where references are included to give arguments the appearance of widespread acceptance, but are actually fake, and/or do not support the argument.
• Falsification is manipulating research materials, equipment, or processes or changing or omitting data or results such that the research is not accurately represented in the research record.
• Plagiarism is the appropriation of another person's ideas, processes, results, or words without giving appropriate credit. One form is the appropriation of the ideas and results of others, and publishing as to make it appear the author had performed all the work under which the data was obtained. A subset is citation plagiarism – willful or negligent failure to appropriately credit other or prior discoverers, so as to give an improper impression of priority. This is also known as, "citation amnesia", the "disregard syndrome" and "bibliographic negligence". Arguably, this is the most common type of scientific misconduct. Sometimes it is difficult to guess whether authors intentionally ignored a highly relevant cite or lacked knowledge of the prior work. Discovery credit can also be inadvertently reassigned from the original discoverer to a better-known researcher. This is a special case of the Matthew effect.
  • Plagiarism-Fabrication - the act of taking an unrelated figure from an unrelated publication and reproducing it exactly in a new publication (claiming that it represents new data). Recent papers from the University of Cordoba have come to light showing how this can go undetected and unchallenged for years.
  • Self-plagiarism – or multiple publication of the same content with different titles and/or in different journals is sometimes also considered misconduct; scientific journals explicitly ask authors not to do this. It is referred to as "salami" (i.e. many identical slices) in the jargon of medical journal editors (MJE). According to some MJE this includes publishing the same article in a different language.

Other types of research misconduct are also recognized:

• The violation of ethical standards regarding human and animal experiments – such as the standard that a human subject of the experiment must give informed consent to the experiment.Failure to obtain ethical approval for clinical studies characterised the case of Joachim Boldt.
• Ghostwriting – the phenomenon where someone other than the named author(s) makes a major contribution. Typically, this is done to mask contributions from drug companies. It incorporates plagiarism and has an additional element of financial fraud.
• Conversely, research misconduct is not limited to NOT listing authorship, but also includes the conferring authorship on those that have not made substantial contributions to the research.This is done by senior researchers who muscle their way onto the papers of inexperienced junior researchers as well as others that stack authorship in an effort to guarantee publication. This is much harder to prove due to a lack of consistency in defining "authorship" or "substantial contribution"

In addition, some academics consider suppression—the failure to publish significant findings due to the results being adverse to the interests of the researcher or his/her sponsor(s)—to be a form of misconduct as well.

• Bare assertions – making entirely unsubstantiated claims - may also be considered a form of research misconduct although there is no evidence that cases of this form have ever led to a finding of misconduct.

In some cases, scientific misconduct may also constitute violations of the law, but not always. Being accused of the activities described in this article is a serious matter for a practicing scientist, with severe consequences should it be determined that a researcher intentionally or carelessly engaged in misconduct. However in most countries, committing research misconduct, even on a large scale, is not a legal offence.

Responsibility of auther  and co authers

Authors and coauthors of scientific publications have a variety of responsibilities. Contravention of the rules of scientific authorship may lead to a charge of scientific misconduct. All authors, including coauthors, are expected to have made reasonable attempts to check findings submitted to academic journals for publication. Simultaneous submission of scientific findings to more than one journal or duplicate publication of findings is usually regarded as misconduct, under what is known as the Ingelfinger rule, named after the editor of the New England Journal of Medicine 1967-1977, Franz Ingelfinger.^[25]

Guest authorship (where there is stated authorship in the absence of involvement, also known as gift authorship) and ghost authorship (where the real author is not listed as an author) are commonly regarded as forms of research misconduct. In some cases coauthors of faked research have been accused of inappropriate behavior or research misconduct for failing to verify reports authored by others or by a commercial sponsor. Examples include the case of Gerald Schatten who co-authored with Hwang Woo-Suk, the case of Professor Geoffrey Chamberlain named as guest author of papers fabricated by Malcolm Pearce,(Chamberlain was exonerated from collusion in Pearce's deception)- and the coauthors with Jan Hendrik Schön at Bell Laboratories. More recent cases include that of Charles Nemeroff,^[28] then the editor-in-chief ofNeuropsychopharmacology, and a well-documented caseinvolving the drug Actonel.

Authors are expected to keep all study data for later examination even after publication. The failure to keep data may be regarded as misconduct. Some scientific journals require that authors provide information to allow readers to determine whether the authors might have commercial or non-commercial conflicts of interest. Authors are also commonly required to provide information about ethical aspects of research, particularly where research involves human or animal participants or use of biological material. Provision of incorrect information to journals may be regarded as misconduct. Financial pressures on universities have encouraged this type of misconduct. The majority of recent cases of alleged misconduct involving undisclosed conflicts of interest or failure of the authors to have seen scientific data involve collaborative research between scientists and biotechnology companies (Nemeroff,Blumsohn).

Responsibilities of research institutions

In general, defining whether an individual is guilty of misconduct requires a detailed investigation by the individual's employing academic institution. Such investigations require detailed and rigorous processes and can be extremely costly. Furthermore, the more senior the individual under suspicion, the more likely it is that conflicts of interest will compromise the investigation. In many countries (with the notable exception of the United States) acquisition of funds on the basis of fraudulent data is not a legal offence and there is consequently no regulator to oversee investigations into alleged research misconduct. Universities therefore have few incentives to investigate allegations in a robust manner, or act on the findings of such investigations if they vindicate the allegation.

Well publicised cases illustrate the potential role that senior academics in research institutions play in concealing scientific misconduct. A King's College (London) internal investigation showed research findings from one of their researchers to be 'at best unreliable, and in many cases spurious'but the college took no action, such as retracting relevant published research or preventing further episodes from occurring. It was only 10 years later, when an entirely separate form of misconduct by the same individual was being investigated by the General Medical Council, that the internal report came to light.

In a more recent case an internal investigation at the National Centre for Cell Science (NCCS), Pune determined that there was evidence of misconduct by Dr. Gopal Kundu, but an external committee was then organised which dismissed the allegation, and the NCCS issued a memorandum exonerating the authors of all charges of misconduct. Undeterred by the NCCS exoneration, the relevant journal (Journal of Biological Chemistry) withdrew the paper based on its own analysis.

Responsibilities of scientific colleagues who are "bystanders"

Some academics believe that scientific colleagues who suspect scientific misconduct should consider taking informal action themselves, or reporting their concerns. This question is of great importance since much research suggests that it is very difficult for people to act or come forward when they see unacceptable behavior, unless they have help from their organizations. A "User-friendly Guide," and the existence of a confidential organizational ombudsman may help people who are uncertain about what to do, or afraid of bad consequences for their speaking up.

Responsibility of journals

Journals are responsible for safeguarding the research record and hence have a critical role in dealing with suspected misconduct. This is recognised by the Committee on Publication Ethics (COPE) which has issued clear guidelines^[35] on the form (e.g. retraction) that concerns over the research record should take.

• The COPE guidelines state that journal editors should consider retracting a publication if they have clear evidence that the findings are unreliable, either as a result of misconduct (e.g. data fabrication) or honest error (e.g. miscalculation or experimental error). Retraction is also appropriate in cases of redundant publication, plagiarism and unethical research.
• Journal editors should consider issuing an expression of concern if they receive inconclusive evidence of research or publication misconduct by the authors, there is evidence that the findings are unreliable but the authors' institution will not investigate the case, they believe that an investigation into alleged misconduct related to the publication either has not been, or would not be, fair and impartial or conclusive, or an investigation is underway but a judgement will not be available for a considerable time.
• Journal editors should consider issuing a correction if a small portion of an otherwise reliable publication proves to be misleading (especially because of honest error), or the author / contributor list is incorrect (i.e. a deserving author has been omitted or somebody who does not meet authorship criteria has been included).

Recent evidence has emerged that journals learning of cases where there is strong evidence of possible misconduct, with issues potentially affecting a large portion of the findings, frequently fail to issue an expression of concern or correspond with the host institution so that an investigation can be undertaken. In one case the Journal of Clinical Oncology issued a Correction despite strong evidence that the original paper was invalid.^[15] In another case,^[36] Nature allowed a Corrigendum to be published despite clear evidence of image fraud. Subsequent Retraction of the paper required the actions of an independent whistleblower.^[37]

The recent cases of Joachim Boldt and Yoshitaka Fujii in anaesthesiology have focussed attention on the role that journals play in perpetuating scientific fraud as well as how they can deal with it. In the Boldt case, the Editors-in-Chief of 18 specialist journals (generally anaesthesia and intensive care) made a joint statement regarding 88 published clinical trials conducted without Ethics Committee approval. In the Fujii case, involving nearly 200 papers, the journal Anesthesia & Analgesia, which published 24 of Fujii's papers, has accepted that its handling of the issue was inadequate. Following publication of a Letter to the Editor from Kranke and colleagues in April 2000, along with a non-specific response from Dr. Fujii, there was no follow-up on the allegation of data manipulation and no request for an institutional review of Dr. Fujii's research. Anesthesia & Analgesia went on to publish 11 additional manuscripts by Dr. Fujii following the 2000 allegations of research fraud, with Editor Steven Shafer stating in March 2012 that subsequent submissions to the Journal by Dr. Fujii should not have been published without first vetting the allegations of fraud. In April 2012 Shafer led a group of editors to write a joint statement,in the form of an ultimatum made available to the public, to a large number of academic institutions where Fujii had been employed, offering these institutions the chance to attest to the integrity of the bulk of the allegedly fraudulent papers.

Photo manipulation

Compared to other forms of scientific misconduct, image fraud (manipulation of images to distort their meaning) is of particular interest since it can frequently be detected by external parties. In 2006, the Journal of Cell Biology gained publicity for instituting tests to detect photo manipulation in papers that were being considered for publication.This was in response to the increased usage of programs by scientists such as Adobe Photoshop, which facilitate photo manipulation. Since then more publishers, including the Nature Publishing Group, have instituted similar tests and require authors to minimize and specify the extent of photo manipulation when a manuscript is submitted for publication. However there is little evidence to indicate that such tests are applied rigorously. One Nature paper published in 2009 has subsequently been reported to contain around 20 separate instances of image fraud.

Although the type of manipulation that is allowed can depend greatly on the type of experiment that is presented and also differ from one journal to another, in general the following manipulations are not allowed:

• splicing together different images to represent a single experiment
• changing brightness and contrast of only a part of the image
• any change that conceals information, even when it is considered to be aspecific, which includes:
  • changing brightness and contrast to leave only the most intense signal
  • using clone tools to hide information
• showing only a very small part of the photograph so that additional information is not visible.

The potentially severe consequences for individuals who are found to have engaged in misconduct also reflect on the institutions that host or employ them and also on the participants in any peer review process that has allowed the publication of questionable research. This means that a range of actors in any case may have a motivation to suppress any evidence or suggestion of misconduct. Persons who expose such cases, commonly called whistleblowers, can find themselves open to retaliation by a number of different means. These negative consequences for exposers of misconduct have driven the development of whistle blowers charters - designed to protect those who raise concerns. A whistleblower is almost always alone in their fight - their career becomes completely dependent on the decision about alleged misconduct. If the accusations prove false, their career is completely destroyed, but even in case of positive decision the career of the whistleblower can be under question: their reputation of "troublemaker" will prevent many employers from hiring them. There is no international body where a whistleblower could give their concerns. If a university fails to investigate suspected fraud or provides a fake investigation to save their reputation the whistleblower has no right of appeal.

With the advancement of the internet, there are now several tools available to aid in the detection of plagiarism and multiple publication within biomedical literature. One tool developed in 2006 by researchers in Dr. Harold Garner's laboratory at theUniversity of Texas Southwestern Medical Center at Dallas is Déjà vu,an open-access database containing several thousand instances of duplicate publication. All of the entries in the database were discovered through the use of text data mining algorithm eTBLAST, also created in Dr. Garner's laboratory. The creation of Déjà vu^[51] and the subsequent classification of several hundred articles contained therein have ignited much discussion in the scientific community concerning issues such as ethical behavior, journal standards, and intellectual copyright. Studies on this database have been published in journals such as Nature and Science, among others

Refernce

https://en.wikipedia.org/wiki/Scientific_misconduct#cite_note-11