To clarify the necessity of applying appropriate statistical methods to calculate sampling error from data of cluster sample survey, we take a two-stage cluster sample survey developed from a population aged 15 and over as example. We use statistical methods based on the assumption of simple random samples and methods considering complex sample design to analyze our data, respectively. Through comparison, we hope to show the potential effects of using improper statistical methods to estimate sampling error on parameter estimation and hypothesis testing. Using standard error algorithms based on the assumption of simple random samples,the standard errors calculated often underestimate sampling error and the hypothesis testing even gets wrong conclusion. When the statistical methods and statistics package for complex survey data are already available, it is necessary for us to use appropriate methods to analyze and report the sampling errors of data from cluster sample survey.
Cluster Analysis ; Data Interpretation, Statistical ; Epidemiologic Studies ; Sampling Studies

OBJECTIVETo estimate the number of death from malignant tumor and to assess the completeness of cancer registry data using the capture-recapture method.

METHODSData about death form malignant tumor were collected during the period of 2004 and 2005 from cancer registry, civil administration and police registries in Linzhou. Cases in every source were matched on the address, name, sex, ID card, date of death, cause of death. A 3-source log-liner poisson model was used to estimate the real number of death from malignant tumor during the period of 2004 and 2005 in Linzhou.

RESULTSThe real number of death from malignant tumor that estimated by capture-recapture method was 3140 (95% CI: 3124 - 3161) during 2004-2005 year in Linzhou. Underreported number by source linkage was 71. Log-liner model demonstrated positive dependence between the 3 sources, implying that 2-source model would underestimate missing cases. Using the mortality number estimated by the capture-recapture as the denominator, under-reporting rate of cancer registry was 6.6%.

CONCLUSIONIt is feasible to estimate the real number of death from malignant tumor and to correct data bias by applying the capture-recapture method.

China ; epidemiology ; Humans ; Neoplasms ; mortality ; Registries ; Sampling Studies

OBJECTIVETo analyze the difference and correlation of food intakes assessed by Food Weighted Record, 24 Hours Recall and Food Frequency Questionnaire (FFQ) methods.

SUBJECTSThe data of 2002 China National Nutrition and Health Survey were used. All 23,198 subjects (male 11,107, female 12,091) finished all three dietary survey methods. Food Weighted Record method weighted all food items consumed at home by all family members in last 3 days; 24 Hours Recall recorded all food items the subjects consumed at and out home in last 3 days; FFQ recorded the main 33 categories food items that the subjects consumed at and out home in last year.

RESULTSCompared to Food Weighted Record method, food intakes assessed by 24 hours recall were similar, and the relative differences were less than 10% of most food items. The proportion of over-report and under-report (+/- 25%) by 24 Hours Recall were less than 40%. The correlation coefficients of food intakes between Food Weighted Record and 24 Hours Recall methods ranged from 0.58 to 0.88. The results from Food Weighted Record and FFQ were similar among the high frequently consumed food items. The relative differences of the food intakes investigated between Food Weighted Record and FFQ were less than 15% and the proportion of over-report and under-report (+/- 25%) by FFQ were more than 50%. The correlation coefficients of food intakes between Food Weighted Record and FFQ ranged from 0.08 to 0.76.

CONCLUSIONFood intakes might be assessed accurately by the Food Weighted Record revised 24 Hours Recall method, and FFQ method be also used to assess the food intakes, especially for the frequently consumed food items, as to studying the relationship between food consumption and health.

Data Collection ; Diet Surveys ; Eating ; Female ; Humans ; Male ; Sampling Studies

No abstract available.
Statistics as Topic ; Models, Statistical ; Data Interpretation, Statistical ; Sampling Studies ; Biometry

OBJECTIVETo study the application of decision tree in the research of anemia among rural children.

METHODSIn the Enterprise Miner module of software SAS 8.2, 3000 observations were sampled from database and the decision tree model was built. The model using decision tree of CART bases on Gini impurity index.

RESULTSThe misclassification rate of decision tree model was, training set 21.2%, validation set 21.9%. The Root ASE of decision tree model was, training set 0.399, validation set 0.404. The area under the ROC curve was larger than the reference line. The diagnostic chart showed that the corresponding percentage was higher than the other. The decision tree model selected 9 important factors and ranked them by their power, among which mother of anemia (1.00) was the most important factor. Others were children's age (0.75), time of ablactation (0.53), mother's age (0.32), the time of egg supplementation (0.26), category of the project county (0.26), the time of milk supplementation (0.16), number of people in the family (0.13), the education status of the mother (0.12). Decision tree produced simple and easy rules that might be used to classify and predict in the same research.

CONCLUSIONDecision tree could screen out the important factors of anemia and identify the cutting-points for factors. With the wide application of decision tree, it would exhibit important application values in the research of the rural children health care.

Anemia ; prevention & control ; Child, Preschool ; Decision Trees ; Evaluation Studies as Topic ; Humans ; Rural Population ; Sampling Studies

OBJECTIVETo investigate level and its influential factors of occupational stressors and strains in workers of a steel plant and explore the feasibility of occupational stress and mental health as indicators of effect evaluation of health promotion at workplace.

METHODS1039 workers were investigated by using cross-sectional method. The questionnaires included job stressors, skill discretion, job control latitude, mental health and sleep disorders.

RESULTS54.5% of respondents reported "job doesn't allow mistakes" as stressor, while 47.6% of respondents reported "too job responsibility" as stressor. The scores of job stressors were significantly higher than female workers (P<0.01). Workers with high education level scored more skill discretion, job control latitude and lower mental health than ones with low educational level (P<0.05 or P<0.01). Elder workers had more skill discretion and job control latitude scores than younger. There were significant differences in scores of skill discretion among different length of service groups (P<0.05). There were also significant differences in scores of skill discretion among different job title groups (P<0.01). Occupational stressors were correlated significantly positively with sleep disorder (r=0.37, P<0.01), but job control latitude negatively to mental health (r=-0.19, P<0.01).

CONCLUSIONThe workers in the steel plant report occupational stressors, which is related to poor self-reported health symptoms. The occupational stressors and mental health should be considered as one of effect evaluation indicators of health promotion at workplace.

Adult ; Burnout, Professional ; Cross-Sectional Studies ; Female ; Humans ; Male ; Middle Aged ; Sampling Studies ; Steel ; Surveys and Questionnaires

OBJECTIVETo explore the relationship between sample size and variance of means for personal noise exposure in weaving workers as to contributing evidence for establishing personal noise exposure measurement guideline.

METHODSA personal noise exposure measurement database from a group of weaving workers was used in the randomized re-sampling data analysis. The sampling cases were one number selecting from one to fifteen at each randomized re-sampling procedure. The randomized re-sampling was one thousand times from original personal noise exposure measurement database to get one thousands of re-sampling database. One thousands of L(Aeq.8 h) mean were calculated by re-sampling databases. The variation of randomized re-sampling means was analyzed for different re-sampling numbers.

RESULTSThe change for narrow trend of maximum, minimum, 95 percent number, 5 percent number of L(Aeq.8 h) mean was faster when randomized re-sampling number was smaller in variation vs randomized re-sampling number curve analysis. After that, the change for narrow trend of L(Aeq.8 h) mean was smooth for increasing the randomized re-sampling numbers. The 95% - 5% of L(Aeq.8 h) mean was about half for randomized re-sampling four cases (3.30 dB) vs one case (7.40 dB), and about one third for seven cases (2.44 dB), and about one fourth for eleven cases (1.85 dB).

CONCLUSIONThe sample size in personal noise exposure measurement guideline could be selected from four to eleven.

Humans ; Noise, Occupational ; statistics & numerical data ; Occupational Exposure ; statistics & numerical data ; Sample Size ; Sampling Studies

Comparative pharmacokinetic (PK) analysis is often carried out throughout the entire period of drug development, the common approach for the assessment of pharmacokinetics between different treatments requires that the individual PK parameters, which employs estimation of 90% confidence intervals for the ratio of average parameters, such as AUC and Cmax, these 90% confidence intervals then need to be compared with the pre-specified equivalent interval, and last we determine whether the two treatments are equivalent. Unfortunately in many clinical circumstances, some or even all of the individuals can only be sparsely sampled, making the individual evaluation difficult by the conventional non-compartmental analysis. In such cases, nonlinear mixed effect model (NONMEM) could be applied to analyze the sparse data. In this article, we simulated a sparsely sampling design trial based on the dense sampling data from a truly comparative PK study. The sparse data were analyzed with NONMEM method, and the original dense data were analyzed with non-compartment analysis. Although the trial design and analysis methods are different, the 90% confidence intervals for the ratio of PK parameters based on 1000 Bootstrap are very similar, indicated that the analysis based on NONMEM is a reliable method to treat with the sparse data in the comparative pharmacokinetic study.
Area Under Curve ; Confidence Intervals ; Humans ; Nonlinear Dynamics ; Pharmacokinetics ; Sampling Studies

Child, Preschool ; China ; epidemiology ; Humans ; Infant ; Infant, Newborn ; Language Development ; Sampling Studies

The purpose of this study was to compare the difference of two survey methods for health services research. Data were collected by means of two types of household survey conducted from March 11 to September 19, 1985. A probability sample of 30,613 persons was taken from 180 Enumerated Districts designated by the Economic Planning Board. The sample was divided into two groups systematically. One group was surveyed by the self-administered questionnaire and the other group was interviewed. Response rates were 81.4% and 90.6% respectively. The data were analyzed by the ratio of the results of the self-administered survey to those of the interview survey. No difference was observed in sex, age, residence, or occupation between the two groups. However the respondents' characteristics were statistically different between the two groups. The major findings of this study are as follows : 1. The morbidity rate was 142.5 per 1,000 persons during the two week period by the self-administered questionnaire survey and 74.3 per 1,000 persons with the interview survey method. The ratio of the morbidity rate by the self-administered questionnaire to that by the interview was 1.92, and the difference between the two rates were due to the personal characteristics. 2. The out-patient utilization rate was 10.2 visits per person per year by the self-administered questionnaire survey and 5.4 by the interview survey, and the ratio was 1.89; the admission rate was 3.2 times per 100 persons per year by the self-administered questionnaire survey and 1.9 times by the interview survey, and the ratio was 1.68. Differences due to the sociodemographic characteristics were greater in the out-patient utilization rates than in the admission rates. 3. Percentages of effective medical care demand were 90.2% in the self-administered survey and 92.3% in the interview survey; the ratio was 0.98 which was less than that of the morbidity rate and medical care utilization. But, differences of effective medical care demand occurred in persons with no occupation, and aged or low educated respondents. 4. Respiratory illness had the highest frequency in the two survey methods. But there was a slight difference between the two survey methods in morbidity composition. 5. It was concluded that data collected by the interview survey were inclined to be underestimated and this problem can be corrected by a modified self-administered survey.
Surveys and Questionnaires* ; Family Characteristics ; Health Services Research* ; Health Services* ; Humans ; Occupations ; Outpatients ; Surveys and Questionnaires* ; Sampling Studies