Error reporting the test statistics and significance levels, and arguable model building

Islam, Md Nazrul (2014) Error reporting the test statistics and significance levels, and arguable model building. Acta Dermato-Venereologica, 94 (2), 252, [Commentary S1].

Record type: Letter

Abstract

Dear Sir,
With academic and professional interests, we read the article by Ekelund et al. (1). Conducted on 443 consecutive patients with psoriasis in Sweden, this paper examined the relationship between measures of disease severity (e.g., as measured by Dermatology Life Quality Index) and associated costs in patients with plaque psoriasis.The authors used linear regression model to examine the relationship between the Dermatology Life Quality Index (DLQI; used as continuous variable; variable name: dlqi_tot), psoriatic arthritis (dummy variable name: dum_psor) and being on systemic treatment (dummy variable name: systemic) in relation to total cost (Table III). All of these three explanatory variables were shown to be statistically signicant, as described by the authors, and depicted by the p-values in the reduced model (Table III).However, the t-values associated with the explanatory variables are too small to reach statistical signicance. It is not clear from the model what the effective sample size was (which might be attributed to missing values etc.), hence the degrees of freedom (DF) could not be denitely identied. However, as the total initial sample size was 443, we may safely assume that the maximum DF would be 439 [DF = n–(p+1); where n = effective sample size, and p = number of predictors (3 in this case) in the model] (2)

This record has no associated files available for download.