DRUG WAR CLOCK - Over $40 Billion per year

Why legalize Hemp/Marijuana? - Economics,
Scientific Evidence. - A key source of info.

A note regarding OBJECTIVITY and the SCIENTIFIC METHOD

As long as a scientist resides in a human body, with all its desires, hungers, wants and strivings, the very thought of bias-free work is impossible. It is only when one eliminates these "defilements" (Buddhist term) that a "Clear Mind" can appear. Since I am far from that point, I am perfectly willing to admit that all of my work contains at least some bias.

One of the most insidious effects of this phenomenon is the selection bias that occurs in determining what is worthy or not worthy of inclusion in research or reporting. DeLoria, in "Red Earth, White Lies" (see below under books) notes the blinders such selection bias creates: "The white man sees only what the white man looks for." I can provide a personal example of the phenomenon. Thus the need for double-blind studies, the idealized form of scientific research. Unfortunately, as Rupert Sheldrake reports, this is rarely practiced even in medical research.

When one really practices the scientific method, all observation and information from whatever source are considered and evaluated to determine the value of and directions that available evidence points. However, few use this paradigm, as evidenced by the following critique from renowned physicist Max Planck: "...a new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it." Here he was discussing the selection bias he faced from those in powerful decision-making positions, who followed, like lemmings, the beliefs they either had been taught or that had made them famous, and who failed to consider contradictory evidence. We clearly see the same in the social sciences, where results based on studies of rats are always given more weight than evidence provided by humans (In fact, any technical report or study, if conducted and written by humans, must, in fact be "anecdotal.").

Myth - The Central Limit Theorem Solves any Problem Resulting from Nonnormal Distributions

Although many fallaciously claim that the Central Limit Theorem solves the problem of nonnormal distributions, this is not really true. Some information on this and the related topic of sample size is provided here. Note that the Lindberg Conditions require a symmetric, light-tailed distribution.

Myth - The Mean-Shift - WOW is this an Absurd Assumption

Parametric statistics assume a shift of means with homogeneous variances for within-group (repeated measures) or between-group tests of significance. Thus they assume that a treatment, or time effect has precisely the same influence on every member of a sample, whether these be rats, humans or beans. Evidence opposed to this assumption should be apparent to anyone who has ever looked at the Physician's Desk Reference (PDR). Every drug affects some individuals greatly and some very little. Some positively, and, in a small percentage of cases, for APA approved drugs, some negatively, even to death. The same is true for any treatment or time effect with any living population (rats, birds, cats, plants, bacteria, etc.). This is one important reason that distribution-free statistics (e.g. rank tests) tend to exhibit greater power to detect differences between groups because they are sensitive to changes in distribution shape in addition to parameter shifts. They are also less influenced by heterogeneity of variance, thus are more robust to alpha.

Is Significance Testing Necessary, or is it too Likely to Mislead the Novice when Large Samples are Available?

There is more than one way to evaluate empirical data. Unfortunately, in many fields, Null Hypothesis Significance Testing (NHST) has come to dominate practice to such an extent that consideration of vital factors such as effect size and meaningfulness has disappeared. A little history is in order. Leading up to today's dogma, first came Student/Gosset's 1908 article on the t and then, over the next 15-20 years, Fisher's development of the widely applicable ANOVA. Despite their frequent disagreements, Fisher adopted Karl Pearson's alpha level arguments for .01, .05 and .10, which have since become almost "set in stone." Over time, journals and academic departments in many disciplines that don't tend to use Bayesian or Likelihood methods, have effectively deified NHST and those signifigance levels (mainly .05). Nix & Barnette following an explication of the Fisherian and Neyman-Pearson approaches to NHST note: "As this brief summary has shown, the simplicity and appeal of the dichotomous decision rule, posited by p values, is alluring. But, it can lead to misinterpretation of statistical significance, and more importantly, it can distract us from a higher goal of scientific inquiry. That is, to determine if the results of a test have any practical value or not."

Are Short-tailed Distributions a Danger?

Much of the robustness literature deals with the detrimental impact that long tails have on parametric statistics. However, due to the bounded/censored nature of many empirical distributions, short tailed distributions occur with some frequency. Hampel notes that this may not be a reason for joy, but for concern.

Is All Validity Really Only "Face Validity"?

Although there are many "types" of validity bandied about in the literature, in fact, everything boils down to whether one or a group of "experts" THINK/FEEL that a particular item or measure contributes or does not contribute to determining the validity of a measure. Thus, it is all based on perceptions, which is basically what Face Validity is about. Obviously, the greater number and types of different sources one has supporting a measure's validity, the stronger the argument one has (sort of like triangulation). This Page discusses validity, while the Site deals with a variety of issues. Campbell & Fiske's Multitrait, Multimethod Matrix is a nice "how to" example of triangulation.

Why Does Measurement Error Propagate in the Physical Sciences, but not in the Social Sciences?

This is more of a plea for sensible use of measurement in all social science fields ranging from Education to Medicine and from Business through Criminology to Psychology. This has a bit of math .