Quota sampling

This chapter presents a partial exploration algorithm guided by a quota sampling. It is basically a table of quotas to fill in during the exploration. The sampling method is neither totally unbiased nor valid variance estimate, but it goes in the direction toward obtaining a representative sample. Then, we show a proper treatment to the sample by means of a hypothesis test [10] to provide conclusiveness to the analysis. 

Non-probabilistic sampling methods appeal more to notions of representativity [14]. In this section we explain how to get a representative sample of the SSp according to the variable Y which is assumed to be normally distributed. The sample is obtained with a partial exploration algorithm guided by a quota sampling [14]. 

In theory, any random sample taken from a normally distributed population should give a normally distributed sample. In practice, we cannot take a random sample from the exploration of the SSp of a PN system. However, by using a quota sampling strategy, we can direct our exploration toward getting the result of a random sampling, a pseudo-normally distributed sample. 

An additional advantage of our analysis method is that, even in the case of stopping the exploration before completing the quota sampling, sometimes is still possible to conduct a hypothesis test with the partial results upon confirmation with a test of goodness of fit that the sample represents a normal distribution. 

Despite the SSp exploration is not a process of randomly generated independent markings, the approach by quota sampling allows us to confirm the highly possible appropriateness of the hypothesis test in order to add completeness to unsuccessful searches, but just in MGPN systems with the three forms presented here. The results... 

G. Goertzel, Quota sampling and importance functions in stochastic solution of particle problems, AECD-2793, December, 1949. 

In this study, the researchers focus on SMEs in the manufacturing sector, the sampled group size was 100 SMEs, and data were collected by quota sampling. With the above sampling groups, a questionnaire was used as a tool for data collection to create a questionnaire checklist by smdying twenty-four variables under all the five performance measurement perspectives (quality, time, finance, and customer satisfaction) for performance measurement of SMEs in the manufacturing sector in the province. 

Quota Sampling n A sampling technique in which a fixed number of elements are chosen from groups of the population. The groups are chosen based on some characteristic of the population usually in an attempt to represent the population as closely as possible. The samples fi-om each group are not necessarily chosen randomly and are often chosen for convenience, which differentiates quota sampling fi om stratified sampling. 

The EDISEN- method allow the compensation of static input capacities Cm 200 pF only by a resistor Rc vs. GND at the input of the Ee102 . This resistor discharge during every sampling cycle a constant quota Cc of the sensor capacity C N.Xo, so the input circuit of the IC has not to discharge Cc. 

The particulate Cd concentrations measured in field samples of surface seawater (Fig. 4) are in the range 0.1-1 mmol CdimolP, equivalent to 1-lOp.mol Cd mol C. They are a good match with the cellular Cd quotas of phytoplankton measured in laboratory cultures under conditions of low (i.e., natural ) metal concentrations which are also in the range 0.1-1 mmol Cd mol P and average 0.2 mmol Cd mol P. The biological utilization of cadmium can thus account quantitatively for the particulate concentration and, hence, for the draw-down of cadmium from surface waters observed in the field. 

In our exploratiOTi, each state might be a vaUd sample. AU samples are segmented into six mutually exclusive groups. The groups are represented in a table of quotas with six slots (slg with g = 1,..., 6). The table of quotas accounts only for n valid samples. Each slot is used as a counter of only valid samples. 

All unique explored markings are registered in a record of states, but only markings which are also entries in the table of quotas are vaUd samples. A marking with cardinality not in the range of any slot of the table is a type-1 not valid sample. A marking with cardinality of a slot with complete quota is a type-2 not valid sample. Repeated markings are discarded as samples. 

The number of type-2 not valid samples is reduced by one every time a valid sample corresponding to an incomplete quota of a slot is found. 

The exploration stops if the target marking is found or when all quotas are complete. In addition, the exploration stops following the same statistical 68-95-99.7 rule for detecting outliers, once the number of type-1 not valid samples is greater than or equal to 0.27% with respect to the tmo-... 

The sample taken with the table of quotas can be seen as a set 4 of markings. The distribution of a variable W, cardinality of the markings in the set 4, appears like a pseudo-normal probability distribution, i.e., ... 

In our reachability analysis, in searching a target marking m, with cardinality of c, the partial exploration algorithm fills in the quota of samples for a table of n markings from the SSp. 

Just for the case of unsuccessful search of the marking m the explored states filling the quotas of the six slots belong to a set 4 of samples. Those samples can be treated with a hypothesis test. 

Table V shows the total Fe (TFe), metallic Fe (Nffe) and metallization rate ( MFe/TFe ) change of the roasted ore after different reduction time periods. It could be seen that TFe kept increase in the initial 100 minutes, and became stable subsequently. The reason is that siderite loses weight through thermal decomposition and reduction while the mass of Fe remain constant. MFe was zero at the beginning since it took time to form the original iron nucleus. Once appeared the MFe grew rapidly, though the acceleration gradually slowed down before it became nearly zero when most of the ferrum was reduced to iron. After 120 minutes when metallization rate peaked at 93.22%, lengthening the roasting time did not make the quotas better. Sample fluctuation was seen in the test, but it did not affect the overall trend.

---