Statistics can help determine whether the difference between the two results is the result of a random phenomenon or it really represents a difference. The statistical way of thinking is: assume that everything happens randomly and see what the result will be. Then the measuremen

statistics can help determine whether the difference between the two results is the result of a random phenomenon or it really represents a difference.

Statistical thinking is: Assuming everything happens randomly, see what the result will be. Then the measurement is done, and if the measurements differ from the expected random results, the observed differences mean that it is not just randomness at work.

The following example can better explain this way of thinking.

A classmate has two lovers, and he cannot choose between them, so he decided to leave the choice for luck.

Two lovers live at both ends of a bus line, and the bus leaves the bus stop in both directions every 10 minutes. He decided to walk randomly at the bus stop and then jump onto the first bus. After 20 days, I walked 16 times to the right and 4 times to the left. He believed that such a result was a clear instruction from God until a friend pointed out that the bus driving to the left was at an hour, 10 minutes after an hour, 20 minutes after an hour... leaving the bus stop, while the bus driving to the right was at an hour, 8 minutes after an hour, 18 minutes after an hour... leaving the bus stop, so the results were completely random, and the difference in the results did not have a significant "statistical" significance. Students have to change other methods to choose.

Therefore, must be reserved and cautious about statistical data.

Statistics in the field of skin care

Generally speaking, statistics are "easy" to apply when a "object" is tested multiple times and it can be assumed that the random error affecting the test is distributed on a Gaussian curve (also known as a "bell-shaped" curve). standard deviation SD represents the "magnitude" of the interval [(mean + SD), (mean – SD)], with approximately 70% of the measured values ​​distributed in this interval range.

Therefore, standard deviation is the index of measurement accuracy .

In skin care trials, when processing tests against individual groups, individual differences increases the difficulty of interpreting the results because it affects the mean and standard deviation before and after the test plan. For example, when the test protocol has an effect on skin hydration, transepidermal moisture loss, skin reflectivity , skin elasticity, or other skin parameters, the group members in the cohort have different values ​​from the outset and will respond differently to the test protocol.

So, the question is: If the data can express the effect, how can we use the data to clarify the effect of the treatment? Has the

test scheme effective?

Usually , testers calculate the standard deviation of the mean sum of the group parameters before and after testing (hydration, elasticity, etc.) and check the difference . Some trialists believe that when the difference in mean values ​​is less than the sum of standard deviations, there is no significant difference in results. other testers, may have called "mysterious" algorithm under the pressure of marketing supervisors to obtain statistically significant judgments.

In both cases, this statistical analysis is far from proof of the actual effect of the test protocol and may also lead to a mask of the therapeutic effect. In fact, trialists should not look at the population data as a whole, they should consider the individual results and calculate the average change . When individuals vary greatly, the experimenter can normalize the results and calculate the average of the percentage changes of each individual in the population. For this clinical test, the relevant measure is not the difference in the population mean before and after the test plan, but the mean of the difference in each participant in the group .

Figure 1 and Figure 2 depict a fictional test result, where the values ​​of the measured parameters are 6, 7, 8, 9, 10, 11, 12, 13, and 14 before the test, and 8, 9, 10, 11, 12, 13, 14, 15, and 16 after the test.

When calculating the mean values ​​M (before) and M (after) and standard deviation, you can obtain M (before) = 10 ±2.5 and M (after) = 12±2.5. Since M(previous)–M(post)=2 and the sum of standard deviations is 5, the difference can be considered insignificant and treatment is ineffective.

Let us consider individual differences: Δ="after"minus "before". The same set of results can be caused by different changes after treatment. Two extreme cases are given in the example of Figure 2: the results can be distributed everywhere (Figure 2, left), or can be characterized by the consistency of each individual in the population in the same direction (Figure 2, right). treatment seems to be completely useless in the first case, while in the second case, the results of the treatment seem to be very meaningful.

on the left and right, the values ​​of "before" and "after" of each group member are connected by a line segment. On the left, the mean value of the difference is Δ=2 and SD=3.3. On the right side, Δ=2, SD=0. When the test results are shown in the figure on the left, you can feel confident that the treatment is ineffective. When the results are shown in the right figure, it is safe to say that the treatment is effective because the mean of the differences is highly statistically significant. Is the effect of

very significant? The statistically significant results of

may actually be completely unrelated.

Consider a paradoxical example, the result of a treatment is that a large number of volunteers can grow two more hairs on the scalp per square centimeters: results may be statistically significant, but in terms of hair growth, it is completely unrelated to .

In this case, statistical applications can beautify the results and allow claiming : This lotion increases hair growth in a statistically significant way. In other cases, as mentioned above, data that showed great efficacy in the trial may instead be discarded without critical analysis of the trial.

Therefore, because of the individual's benchmark differences, the difference in response to the test, and the time required for skin care testing (usually several weeks), the only way to evaluate the test effect is to analyze individual differences . When using statistical methods to strictly process these data, the mean value of the deviation should be considered, rather than the deviation of the mean value.

Author of this article: Paolo Giacomoni, Ph.D.

Paolo Giacomoni, Ph.D., is an independent consultant in the skin care industry. He has served as the Executive Director of Research at Estee Lauder and the head of L'Oreal's Biological Department. He has made a series of achievements in the research on DNA damage and metabolic disorders caused by ultraviolet radiation, as well as the positive effects of vitamins and antioxidants. He has written more than 100 peer-reviewed publications and owned more than 20 patents.

This article is translated from "Happi" magazine

Author: PaoloGiacomoni

Compilation and layout: Susie / John

Original statement:

All original content on this website is prohibited from reprinting or excerpting without permission, otherwise Jung Industrial Media reserves the right to hold accountable. For any platform that has reprinted original articles on this website without permission, please delete the relevant articles immediately.

For more industry information, you can follow the "Jung Personal Care" official account, or log in to Jung Industrial Resources Network https://www.industrysourcing.cn/personalcare to learn about it.