Größe Mittelwertunterschiede Drosten-Kontroverse |
| 29.05.2020, 19:00 | DanielBerlin | Auf diesen Beitrag antworten » |
| Größe Mittelwertunterschiede Drosten-Kontroverse In der Studie werden die Mittelwerte für den Basis-10-Logarithmus der Viruslast angegeben: „The mean (…) are shown for the base-10 logarithm of viral load“. In einem anderen Artikel heißt es: „Results are expressed as the log10 of estimated number of viral particles per ml“. Die Kindergarten-Kinder haben den Mittelwert 4,4, die Erwachsenen den Mittelwert 5,2. Ich frage mich, ob dieser Unterschied klein oder groß, unbedeutend oder bedeutend ist? Die Werte 4,4 und 5,2 klingen für sich genommen nicht nach einem Riesenunterschied, sondern eher nahe beieinanderliegend . Aber es geht ja auch um den Basis-10-Logarithmus. Herr Drosten erweckt in Interviews immer noch den Eindruck, als wenn das zentrale Ergebnis seiner Studie eigentlich stimmen würde, nur der Signifikanz-Test sei nicht optimal gewesen. Stimmt das, dass die Viruskonzentration der untersuchten Kinder und Erwachsenen ungefähr gleich ist, oder ist das eine Nebelkerze? Sollte es so sein, dass die untersuchten Kindergarten-Kinder wesentlich weniger Viruslast hatten als die untersuchten Erwachsenen, dann hätte Christian Drosten die Studie längst für falsch erklären und sie zurückziehen müssen. Vielen Dank im Voraus! Noch einige ergänzende Informationen aus verschiedenen Papern: Leonhard Held schreibt in seinem Paper „A discussion and reanalysis of the results reported in Jones et al. (2020)“: „The authors argue convincingly that the non-normal distribution of the outcome does not allow application of standard parametric statistical procedures for the original data. However, a test for trend is still possible based on the mean log10 viral loads (and standard errors) reported in Table 2 of the paper (assuming normality of the means as justified by the central limit theorem). I have conducted such an analysis using meta-regression techniques and the results show some evidence for an increase of mean log10 viral load with age group. The estimated linear relationship is shown as a red line in Figure 1 and 2. The p-values of the test for trend are 0.047 and 0.015 for categorization C1 and C2, respectively. Both slope estimates are positive, suggesting a positive association between age and viral load. The increase in mean log10 viral load for one age group category increase is estimated to be 0.045 in categorization C1, which corresponds to a 10.8% increase in viral load per 10 year increase in age (95% CI from 0.1% to 22.6%). This range seems to cover clinically relevant differences between age groups.“ „A reanalysis of summary data with a test for trend suggests that there is moderate, but not overwhelming evidence for increasing viral load with increasing age.“ Dominik Liebl schreibt in seinem Paper „OPEN REVIEW OF THE MANUSCRIPT: AN ANALYSIS OF SARS-COV-2 VIRAL LOAD …“: „The mean viral load of the age-group Kindergarten is by 86% lower than the mean viral load of the age-group Mature.“ „3.1. Unnecessary log-transformation. The authors transform the original data using a logarithmic transform to base 10. The non-parametric, rank-based tests used in Jones et al. (2020), however, are invariant to such a logarithm transformation. Thus, the data could also be analyzed on the original scale. This would simplify the interpretation of the significant differences in the virus loads.“ „3.2. On the magnitude of the significant difference. Table 1 shows the means of the two significantly different groups of categorization C2 as reported in Table 2 of Jones et al. (2020), but back-transformed to their original scale (viral copies per mL). The mean viral load of the age-group Kindergarten is by 86% lower than the mean viral load of the age-group Mature.3“ Fußnote 3: 86.1348% = (10 5.229369 – 10 4.371295)/10 5.229369 „Remark. Since non-parametric, rank-based tests procedures are used to analyse the data, one should present the medians - not the means. Moreover, the unnecessary logarithmic data-transformation of the data complicates a proper assessment of the differences in the means (see Jensens’s inequality).“ David Curtis schreibt in seinem Paper „Children have lower SARS-CoV-2 viral loads than adults“: „Given the measurements on the log scale, the best estimate of the magnitude of this difference is that it represents an approximately threefold difference in measured concentrations between children and adults with detectable viral RNA. However one should bear in mind that some subjects may have had viral concentrations below the detectable level and one can note that a smaller proportion of children had detectable virus than adults.“ Kevin McConway und David Spiegelhalter schreiben in ihrem Paper „Is SARS-CoV-2 viral load lower in young children than adults? Jones et al provide evidence that it is (in spite of their claims to the contrary).“: „It has been widely reported as implying that viral loads in children are similar to adults, and yet the data in the article show children between 1 and 10 having on average 27% (conservative 95% interval 8% to 91%) of the viral load of adults aged over 20.“ |
||
|
|
Doppelpost!