The AstroStat Slog

There were (only) four articles from ADS whose abstracts contain the word semiparametric (none in titles). Therefore, semiparametric is not exactly [MADS] but almost [MADS]. One would like to say it is virtually [MADS] or quasi [MADS]. By introducing the term and providing rare examples in astronomy, I hope this scarce term semiparametric to be used adequately against its misguidance of astronomers to inappropriate usage for statistical inference with their data.

• [2006MNRAS.369.1334S]: semiparametric technique based on a maximum likelihood (ML) approach and Voronoi tessellation (VT). Besides, I wonder if Section 3.3, the cluster detection algorithm works similarly to a source detection algorithm in high energy astrophysics if tight photon clusters indicate sources. By the way, what is the definition of sources? Depending on the definitions, determining the right thresholds for detections would change; however, it seems like (brute) Monte Carlo simulations i.e. empirical approaches are employed for setting thresholds. Please, note that my questionnaire is irrelevant to this paper, which I enjoyed reading very much.
• [2004MNRAS.347.1241S]: similar to the above because of the same methodology, ML, VT, and color slide/filter for cluster detection
• [2002AJ....123.1807G]: cut and enhance (CE) cluster detection method. From the abstract: The method is semiparametric, since it uses minimal assumptions about cluster properties in order to minimize possible biases. No assumptions are made about the shape of clusters, their radial profile, or their luminosity function. On the contrary, I wish they used nonparametric which seems more proper in a statistical sense instead of semiparametric judging from their methodology description.
• [2002A%26A...383.1100N]: statistics related keywords: time series; discrete Fourier transform; long range dependence; log-periodogram regression; ordinary least squares; generalized least squares. The semiparametric method section seems too short. Detail accounts are replaced by reference papers from Annals of Statistics. Among 31 references, 15 were from statistics journals and without reading them, average readers will not have a chance to understand the semiparametric approach.

You might want to check out wiki:Semiparametric about semiparametric (model) from the statistics standpoint.

The following books that I checked from libraries some years back related to semiparametric methods, from which you could get more information about semeparametric statistics. Unfortunately, applications and examples in these books are heavily rely on subjects such as public health (epidemiology), bioinformatics, and econometrics.

• Rupert, Wand, and Carroll (2003) Semiparametric Regression, Cambridge University Press
• HУ?rdle, MУМller, Sperlich, and Werwatz (2004) Nonparametric and Semiparametric Models, Spinger
• Horowitz (1998) Semiparametric Methods in Econometrics (Lecture Notes in Statistics) , Springer

There seem more recent publications from 2007 and 2008 about semiparametric methods, targeting diverse but focused readers but no opportunities for me to have a look on them. I just want to point out that many occasions we confront that full parametrization of a model is not necessary but those nuisance parameters determines the shape of a sampling distribution for accurate statistical inference. Semiparametric methods described in above papers are very limited from statistics viewpoints. Astronomers can take a way more advantages from various semiparametrical strategies. There are plenty of rooms for developing semiparametric approaches to various astronomical data analysis and inference about the parameters of interest. It is almost unexplored.