In the case of questioning character flaws, yes, because it is hard to admit those, especially to ourselves. I do think, though, that if Socionics information elements exist, it is possible to create standardized tests for them other than self-report questionnaires. We did a similar thing regarding mental rotation. Psychologists wondered how people conclude that two figures are the same, but rotated, or mirrors that do not fit onto each other. They hypothesized that people would mentally rotate the images in their head to see if they overlap, just like we would do with real figures. The null hypothesis, on the other hand, stated that people would immediately grok the differences. If the null hypothesis was true, this meant that people would not take longer for solving some figures compared to others. If the alternative hypothesis was true, however, people would take longer to solve a figure the more it was rotated away from the other, because they had to rotate it back in their mind, until the two overlapped. This latter explanation turned out true. It is one of my most favourite applied cognitive psychology experiments, because it makes invisible cognitive processes visible even before the advent of advanced brain scanners, such as fMRI.Originally Posted by Stray Cat
That instead sounds like centering, which is substracting the mean of all the respondents for that variable of each individual score. For instance, let's say you use a questionnaire item called "Love of dogs" of which respondents can report their disagreement or agreement on a scale ranging from 1 to 7. Assume that most respondents feel slightly positive towards dogs, if only just for all the cute puppies there are, so on average people score around 4,2 on our "Love of dogs" item. We then substract this from each individual respondent's answer. So a respondent who scored 5 would turn into one who scored 0,8, while one who scored 2 now scores -2,2. The closer people score towards the average, the closer their new score is to 0. A respondent with a score of 4 would thus become one with a score of -0,2. This way it becomes easily observable if respondents scored below or above the mean. It also helps to normalize the data, which is a fancy term for some criteria data have to fulfill, in order to be properly analysable.
Most data sampled from large respondent groups is normally distributed by default, but sometimes you have to help the data a bit. Through centralization we can achieve this. The relative differences between the scores are retained, as the difference between respondent A and B is the same regardless of if you use the old values of 5 and 2, or the new ones of 0,8 and -2,2, because the difference remains 3 in either case. However, the original value of the scores is lost. In this case of something abstract like "Love of dogs" on which 1 up to 7 are not concrete values, it doesn't matter, but if we instead measured the number of dogs that one owns, instead of their love of them, this suddenly would become an issue. What does a centralized score of -3.2 dogs mean? It is hardly interpretable.
When centralizing doesn't show to be sufficient, statisticians go the next step and standardize the data, in order to make it conform to normality. This we do by dividing the new, centralized scores by the standard deviation. The standard deviation is the total difference of each score with the mean. This total difference we then divide by the total number of respondents, so we get the mean total difference of the scores, which we call the variance. If you take the square root of the variance, you obtain the standard deviation. If you standardize data, they basically always come to fall onto the standard distribution and you can analyse them, because the standard distribution is normally distributed. But now the absolute values no longer interpretable, so the -3,2 dogs, but also the relative differences are lost. Respondents A and B no longer differ 3 dogs from each other, but a number of standard deviations from one another. With each transformation even less normally distributed data start behaving normally and thereby become analysable, but at the expense of interpretability.
Yes, if that negative correlation is significant, then we conclude that "Introversion and dog loving are negatively correlated, which instead of showing that introverts like dogs, they actually dislike them." But if the correlation is too weak and thus non-significant we instead say: "We retain the null hypothesis of there being no relationship between introversion and dog loving."
In either case we cannot jump to the conclusion that introverts are cat lovers, because we did not examine this. Yes, most dog lovers prefer dogs over cats and most cat lovers prefer cats over dogs, but it would be possible that introverts dislike both dogs and cats. We would therefore have to do another study relating introverts with cat loving, then we can say something about introverts being cat lovers, but based on the dog loving study alone we cannot conclude this.
Yes and yes, you're getting it.
"If personality fits onto continuous scales" It does actually, that's what the whole Big Five Personality Test is after all, five continuous scales that people can score on.
", with people not deviating much from the mean," That's the very definition of the mean, the average score, so most people barely deviate from it, because otherwise it would not be the mean. Or you have some influential outliers. Let's take the average American income, which is an okay $71.456,- ( Statista, 2021, https://www.statista.com/statistics/...yee-in-the-us/ ). But now we exclude all the outliers, so all the Jeff Bezoses, Elon Musks, and Mark Zuckerbergs from the equation. They namely differ substantially from the rest of the population, because the eight richest people own as much as half the poorest of humanity.When you do this, the mean suddenly drops and you find out that the average American actually earns less than $71.456,-. So the $71.456,- is nice for politicians to claim the average American makes. When the rich become richer, the politicians can truthfully claim that the average American income has gone up without the common man earning a single dime more in reality.
Excluding outliers isn't that desirable either, though, because one study can exclude only the eight richest Americans from the average income, but another could filter out the entire Forbes Billionaires list. The two would thus reach different average American incomes based on these decisions. This is why we take the median when we analyse data prone to many influential outliers, such as income. The median is the middle score. If we have 1.001 respondents and we sort their incomes from small to large, the median income would be the middle one, so the 500th one. Although it is not as easy and informative to calculate with the median, instead of the average, in the case of outlier prone data it is well worth the trade off.
Then there also exists the mode, which is the number that pops up most frequently. Like the median it is impervious to outliers, yet it can still provide different results than the median. Imagine a "fictional situation" of the United States of America being split in a lower and an upper class. The elite are at the top of the pyramid, whereas the lower class is composed of working men and women. If the data set is hetereogeneous the variance of the lower class is for instance much less, than that of the upper class. This means that most workers earn roughly the same, whereas wealth has more variance, because one can be Ferrari rich, personal yacht rich, or private space rocket rich. If you take the mode, which is the most reoccuring number, then you would take the wage of any of the many common workers. But if you take the median, you might instead end up with a no student debt moderately rich person. In addition, many analyses can work with the median, but not the mode.
It thus all depends on what data one examines and with what purpose that determines the choice of statistical instruments, but if possible researchers will use the mean and if the data are prone to influential outliers they use the median.
The political compass functions according to similar principles, though with only two axes and fewer categories. https://pbs.twimg.com/media/EZmZDQFX...png&name=large We have a libertarian-authoritarian scale and a state controlled-market controlled scale. There are many data points of which more are centered around the middle than farther away. Still, one can classify different categories. I use the European Parliament as an example, because it actually does have multiple categories. The American system only has two parties in congress, which might mislead one to believe that one could only deduce two categories based on four axes, which is wholly untrue and merely a quirk of the Anglo-Saxon system.
If we look at the spread of the data the national-conservative European Conservatives and Reformists (ECR), Identity and Democracy, and the non-inscrits share being the most authoritarian of all parties and are pro-market.
The centre-right European People's Party (EPP) and unaffiliated representatives are less authoritarian, but as right-wing as the national conservative ECR and the non-inscrits.
The liberal Renew Europe (RE) is distinguishable as the only libertarian pro-market party in the European Parliament.
The centre-left Socialists and Democrats (S&D) are, as their title implies, centre-left and are thus slightly pro-state. They are also slightly libertarian.
The Greens-European Free Alliance (Greens/EFA) are as pro-state as the S&D, but even more libertarian.
The left-wing Gauche Unitaire Européenne/Gauche Verte Nordique (GUE/GVN) is spread in libertarianism amongst the S&D and Greens/EFA, but are more pro-state than either of them.
Just based on these data points without any colour differences and legend I would be able to recategorize these data points based on their position on these two scales. I might categorize some S&Ds as being unaffiliated, and I would be unable to distinguish all those three authoritarian pro-market parties, but other than that I would be pretty accurate at reclassifying them. I actually created an algorithm that had to classify iris flowers into three different sub-species based on four axes, namely petal and sepal length and width. This clustering algorithm was able to quite accurately categorize the data when comparing its output to the actual sub-species labels of the flowers. It illustrates that it thus is possible to categorize data based on continuous scales, even when the differences are rather small. Here are the flowers: https://s3.amazonaws.com/assets.data...nelearning.png
And this is what some of the data looks like: http://www.sthda.com/english/sthda-u...and-size-1.png
The more significant scales you have, the more accurately you will be able to classify groups even when they differ very slightly from one another. But even with two axes it is already possible to relatively reliably create several categories, as illustrated with the political compass. It is thus possible to distinguish most ESIs and EIIs from one another, when provided enough significant scales. But even with few scales it is already feasible to differentiate ESIs and ILEs from one another, which I think the ESIs can confirm, right @Lady Lunacik, @wonderwoman, @Suonani?
Personality disorders are defined as extreme personality traits, so by definition it would only pertain to a minority of people. But having personality in and of itself is not exclusive to a minority, as all people have personality. Socionics is about personality and is thus applicable to the majority of people, not the minority.
Reply With Quote