Problem of small sample size

Support Vector Machine Data Processing Method for Problems of Small Sample Size ... 

Problem of small sample size is a technical term used by Vapnik. It means a problem of small data set. 

Since many computerized prediction problems involved in this intelligent database are problems of small sample size, support vector machine plays an important role in the model-building for this intelligent database [7]. 

In the semi-empirical method of molecular design, we have to use the data of known molecules as the training set. In many cases, the numbers of available data of known molecules used in the training sets are rather limited. Therefore the data processing tasks are usually dealing with the problem of small sample size, so support vector machine should be suitable for solving these problems. 

The ways to eliminate variability in quantitative work using Py-GC/MS include the use of small sample size and of instruments with the temperature profile well calibrated. The analysis of a standard polymer at specified intervals of time (or number of samples) helps to verify reproducibility, and careful evaluation of the factors that may influence variability followed by their elimination may improve the results. One other procedure to improve reproducibility is the use of an internal standard during pyrolysis. The addition of a standard is not always simple when the sample weight is below 1 mg, and the standard must represent only a small part of the sample. A solution to this problem is to pyrolyze simultaneously with the sample a measured amount of standard diluted in a solid matrix, an example being alumina containing 1% of 1,4-dibromobenzene [40]. 

Sampling and sample preparation are cornerstones of any analytical methodology. Probably the most important advances in this field are miniaturization, simplification of methodologies, and their adaptation to HT. On the technical side, using devices analogous to very simplified chromatography have become widespread. These usually take the form of a small disposable cartridge, such as SPE, ultrafil-tration tubes, etc. These are very efficient, are easy to adapt to the required problem, need small sample size, minimize problems related to contamination (carryover), and are easy to automate. Most other sample preparation equipments serve to support or complement these techniques. 

The presence of noise gives rise to many difficulties in the data processing work in chemistry and chemical technology, especially in the problems with small sample size. 

For samples with low values of K, < 500 ppm, the laser method, because of its small sample size, becomes more uncertain than a thermal release study with a larger sample size. It is not possible to compensate in the laser probe method by multiple pulses above MOO individual pulses which corresponds to evaporation of approximately 20 pg. It is conceivable that by completely redesigning a mass spectrometer with smaller volumes which is bakeable to higher temperatures to alleviate this problem. We have been able on occasion to work with samples as low as 100 ppm K but certainly not on a routine basis. 

Small sample sizes are often required to minimize thermal conductivity problems. Less satisfactory than DSC with regard to resolution of thermal traces and quantitative data. 

Small sample size and sample introduction are significant problems with CE. Furthermore, CE is unable to offer the range of selective separations, including preparative separations, that are possible with HPLC. 

Finally, Stroberetal. (183) conducted an 18-month, prospective, naturalistic follow-up study of 37 bipolar-l adolescents (i.e., 13 to 17 years old) stabilized on lithium and found a relapse rate almost three times higher in those who discontinued prophylactic lithium (92%), as compared with those who complied (38%). Further, they noted that earlier relapse in these patients predicted a greater risk of subsequent relapse and that an early onset may be associated with a more virulent course, resistance to lithium, and the need to consider alternative mood stabilizers. Methodological problems with this study included a small sample size lack of assessment for personality disturbances and intrafamilial environment only a 4-week initial drug stabilization period and lack of precision in monitoring compliance to treatment. 

SPEs offer distinctive advantages over conventional liquid-liquid extractions. They are relatively fast, require small sample size, eliminate emulsification problems, provide the possibility of performing both cleanup and preconcentration of the extract in one analytical step, and offer high precision. Another great advantage of SPEs over liquid-liquid extractions is solvent savings. Unlike liquid-liquid extractions that often require hundreds of milliliters for single or multiple extractions, SPEs require only a few milliliters of solvents for analyte extraction and cleanup. 

---