Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Linear Regression”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 577 records · Page 32Linked to original sources
PubMed1989

Computerized curve fitting in the analysis of hydrogen gas clearance curves.

Hydrogen gas clearance curves obtained from the rat gastric corpus were digitized into a computer and then analyzed by three methods: 1) linear regression of log-transformed data, 2) direct curve fitting with a modified Gauss-Newton nonlinear regression algorithm, and 3) Zierler's height-over-area algorithm. For linear regression of log-transformed data, if the initial base-line estimate was inaccurate or normal amounts of experimental noise were present, the log-transformed data was skewed, leading to deviation of the regression line and incorrect estimation of blood flow. By utilization of the direct-fit routine, the initial estimate of the parameters or experimental noise had little influence on the blood flow determination because of iterative improvement of the parameters. In a study of isoproterenol-stimulated gastric blood flow, Zierler's algorithm underestimated the blood flow estimate. We conclude that analysis of hydrogen gas clearance curves by linear regression of log-transformed data or by Zierler's algorithm may potentially introduce errors in blood flow estimates that may be avoided by analysis with a direct-fitting, nonlinear regression algorithm.

Algorithms↗
PubMed1997

Prophylactic transabdominal amnioinfusion in oligohydramnios for preterm premature rupture of membranes: increase of amniotic fluid index during latency period.

OBJECTIVE: This study was designed to: (i) evaluate the effect of amnioinfusion on the latency period in patients with oligohydramnios for preterm premature rupture of membranes, and (ii) to investigate the relationship between changes in the amniotic fluid index and fetal heart rate short-term variability by computerized Hewlett-Packard cardiotocography, longitudinally estimated before and after prophylactic amnioinfusion. MATERIALS AND METHODS: All singleton pregnancies with prolonged premature rupture of membranes after 25 weeks of gestation and seen at the Institute of Obstetrics and Gynecology, University of Ancona (Italy), between January 1994 and June 1995 were included in the study. Transabdominal amnioinfusion with 150-350 ml warmed normal saline (25-50 ml/min) was performed at weekly intervals. Amniotic fluid volume was assessed ultrasonographically by means of the four-quadrant technique on a weekly basis before and after each amnioinfusion, as well as the short-term variability by a Hewlett-Packard computerized cardiotocographic system. RESULTS: 18 women were enrolled and underwent prophylactic transabdominal amnioinfusion at weekly intervals until delivery. Eighteen controls, who did not undergo prophylactic amnioinfusion, were recruited from our 1992-1993 series and included in the study. The median interval between premature rupture of membranes and delivery was 3.0 weeks (range 1-8 weeks), with an average delivery age of 33.0 weeks (range 27-36 weeks). The latency period was significantly longer in patients who underwent prophylactic amnioinfusion (mean +/- SD, 4.1 +/- 1.7 weeks) than in controls(1.7 +/- 1.0 weeks; p < 0.001). An increase in both the weekly amniotic fluid index (linear regression analysis r = 0.8, p = 0.03) and the weekly short-term variability (linear regression analysis r = 0.82, p = 0.02) was observed among patients who underwent prophylactic amnioinfusion. A direct relationship was observed between the amniotic fluid index and short-term variability (linear regression analysis r = 0.54, p = 0.04). The mean values of fetal movements recorded by computerized tomography during the 20 min of observation significantly increased after amnioinfusion in comparison with those before it (2.6 +/- 0.9 and 0.9 +/- 0.7 respectively; p = 0.001). CONCLUSION: The present study has shown a positive effect of prophylactic transabdominal amnioinfusion on the latency period in patients with preterm premature rupture of membranes and oligohydramnios. Among the patients who underwent amnioinfusion, an interesting improvement in fetal heart rate short-term variability was associated with the progressive increase in amniotic fluid volume, as an expression of fetal well-being.

Abdomen↗
PubMed2004

Multivariate spectral calibration technique based on regression analysis for the quantitative multiresolution of a ternary mixture containing paracetamol, ascorbic acid and acetylsalicylic acid in effervescent tablets.

Multivariate spectral calibration techniques based on regression analysis were established for the quantitative multiresolution of a ternary mixture containing parecetamol (PAR) ascorbic acid (AA) and acetylsalicylic acid (ASP) having closely overlapping spectra. The mathematical algorithms of multivariate spectral calibrations as namely tri-linear regression calibration (TLRC) and multi-linear regression calibration (MLRC) are based on the use of the linear regression equations at a three-wavelength set and a ten-wavelength set in the range of 215-305 nm. These calibration techniques do not require any chemical pre-treatment and a graphical procedure of the overlapping spectra. The mathematical content of TLRC and MLRC approaches were briefly formulated for the quantitative analysis of three- or multi-component mixtures. The applicability of the formulated calibration models were tested by analysing the various synthetic ternary mixtures consisting of these active compounds and then these models were applied to real pharmaceutical formulations. It was observed that TLRC and MLRC models give a successful quantitative multiresolution. The experimental results of these techniques were compared with each other as well as with those obtained by literature methods.

Acetaminophen↗
PubMed2001

A comparison of limited sampling strategies for prediction of Ecteinascidin 743 clearance when administered as a 24-h infusion.

PURPOSE: Ecteinascidin 743 (ET-743) is a novel, marine-derived anticancer agent currently under clinical development for the treatment of solid tumors. The aim of this study was to develop and validate limited sampling strategies for the prediction of ET-743 clearance in phase II studies, using two techniques: the stepwise linear regression approach and the Bayesian estimation approach. METHODS: Data from a phase I dose-finding study were used with ET-743 administered as a 24-h infusion. Plasma concentration time data from 34 patients treated with 1200. 1500 or 1800 microg/m2 ET-743 were randomly divided into an index data set, used for the development of the strategies, and a validation data set. With the linear regression approach, clearance (obtained by non-compartmental analysis) was correlated with the ratios of dose to the observed concentrations. For the Bayesian approach a three-compartment population pharmacokinetic model was developed; optimal time-points were selected using the D-optimality algorithm. The strategies were compared by assessment of their predictive performance of CL in the validation data set. RESULTS: The linear regression method yielded a single-point sampling schedule with no significant bias and acceptable precision (-0.03% and 21%, respectively). With the Bayesian approach, a three-sample strategy was selected which resulted in less-accurate, but unbiased, predictions (bias 13%, precision 34%). CONCLUSIONS: Optimal sampling strategies were developed and validated for estimation of ET-743 clearance. Although the linear regression approach showed slightly better predictive performance, the Bayesian approach is preferred for the current phase II studies as it is more robust and flexible and allows the description of the full pharmacokinetic profile.

Adult↗
PubMed1976

A review of the satatistical considerations involved in the treatment of isotope dilution calibration data.

The use of linear regression analysis for the reduction of isotope dilution data is reviewed. The calculation of linear regression statistics is based upon four assumptions: zero variance in the independent variable, equal variance for all values of the dependent variable, linearity and continuity. Unfortunately, isotope dilution data often violate one or more of these assumptions, which results in the calculation of an inaccurate calibration line. The inaccuracies can be avoided through careful inspection of the data, including analyses of variance and linearity. Large differences in the variances of the dependent variable require the use of a weighted linear regression. Nonlinearity necessitates either discarding data in the nonlinear portion of the calibration or the calculation and use of atom % excess and dilution instead of the sample isotope ratios.

Mass Spectrometry↗
PubMed1999

Evaluation of different weight correction methods for antenatal serum screening using data from two multi-centre programmes.

Weight correction of serum markers is widely used when screening for Down's syndrome and open neural tube defects (NTD) because marker concentrations decrease with increasing maternal weight. Log-linear regression is frequently used for determining weight correction factors, but recently reciprocal-linear regression has been suggested to have advantages. We compared both methods of weight correction using data from two screening programmes carried out by this laboratory, one using alpha-fetoprotein (AFP) and total human chorionic gonadotrophin (HCG) (n = 129,143) and the other, AFP and free beta-HCG (n = 39,982). The reciprocal-linear method fitted the data more closely but did not significantly alter the detection rate or screen positive rate (SPR) for Down's syndrome or NTD with either dataset. Without correction, women heavier or lighter than average weight had significantly different SPRs for Down's syndrome and NTD compared with those weighing close to the median weight. Both correction methods smoothed out the variability in the SPR for Down's syndrome to a similar degree, but reciprocal-linear regression was much better at reducing the variability in SPR for NTD and its use is therefore worthwhile.

Biomarkers↗
PubMed2002

Using continuous PM2.5 monitoring data to report an air quality index.

As stated in 40 CFR 58, Appendix G (2000), statistical linear regression models can be applied to relate PM2.5 continuous monitoring (CM) measurements with federal reference method (FRM) measurements, collocated or otherwise, for the purpose of reporting the air quality index (AQI). The CM measurements can then be transformed via the model to remove any bias relative to FRM measurements. The resulting FRM-like modeled measurements may be used to provide more timely reporting of a metropolitan statistical area's (MSA's) AQI. Of considerable importance is the quality of the model used to relate the CM and FRM measurements. The use of a poor model could result in misleading AQI reporting in the form of incorrectly claiming either good or bad air quality. This paper describes a measure of adequacy for deciding whether a statistical linear regression model that relates FRM and continuous PM2.5 measurements is sufficient for use in AQI reporting. The approach is the U.S. Environmental Protection Agency's (EPA's) data quality objectives (DQO) process, a seven-step strategic planning approach to determine the most appropriate data type, quality, quantity, and synthesis for a given activity. The chosen measure of model adequacy is r2, the square of the correlation coefficient between FRM measurements and their modeled counterparts. The paper concludes by developing regression models that meet this desired level of adequacy for the MSAs of Greensboro/Winston-Salem/High Point, NC; and Davenport/Moline/Rock Island, IA/IL. In both cases, a log transformation of the data appeared most appropriate. For the data from the Greensboro/Winston-Salem/High Point MSA, a simple linear regression model of the FRM and CM measurements had an r2 of 0.96, based on 227 paired observations. For the data from the Davenport/Moline/Rock Island MSA, due to seasonal differences between CM and FRM measurements, the simple linear regression model had to be expanded to include a temperature dependency, resulting in an r2 of 0.86, based on 214 paired observations.

Air Pollutants↗
PubMed1997

Measurement of infant pulmonary mechanics: comparative analysis of techniques.

In spontaneously breathing, intubated infants, lung elastance (E) and resistance (R) can be derived from the general equation of motion of the lung, Ptp = E.V + R.V + I.V + K, where P is the pressure applied to the system, E is the elastance (the reciprocal of the compliance), V is the volume, R is the resistance to airflow (V), I is the inertance, (V) is gas acceleration, and K is a constant representing alveolar pressure. However, differences in the methods of analysis may results in variations in the derived values of E and R. Eighteen non-sedated intubated infants, with a mean post-conceptual age of 32.3 weeks and mean weight of 1.51 kg, were studied to compare values in E and R derived by 7 different methods of analysis. For each infant, the same 7 consecutive breaths were analyzed by the following methods: 1) linear regression of Ptp on tidal volume (VT); 2) a Mead-Whittenberger analysis, 3) a 2-factor linear regression of Ptp with (V) and VT; 4) a 3-factor regression of Ptp with V, VT, and V; 5) a 3-factor linear regression model of Ptp with V, VT, and a (V)2 factor (to incorporate a turbulent flow factor); 6) a 4-factor regression of Ptp with V, VT, V, and (V)2; and 7) a transformation regression plot of Ptp/VT versus V/VT. In addition, the lung impedance was calculated from the values of E and R derived by each method. There was no significant difference between the measurements of E and R in methods 2-6 (by MANCOVA, with the subjects being covariates). Methods 1 and 7 had a significantly elevated E; in addition, Method 7 also had an elevated R. Furthermore, there were no differences between the impedance values derived from the different methods. We conclude that in this group of spontaneously breathing, intubated infants with a spectrum of pathology, the results of a 2-factor regression analysis are comparable to more complex models of the mechanics of the lung.

Airway Resistance↗
PubMed2006

Correlation between image noise and body weight in coronary CTA with 16-row MDCT.

RATIONALE AND OBJECTIVES: To evaluate the correlation between image noise and body weight (BW) or body mass index (BMI) in coronary computed tomography angiography (CTA) as a potential parameter for reducing radiation dose in coronary CTA. MATERIALS AND METHODS: Thirty-six patients who underwent electrocardiogram-gated cardiac CT were analyzed in this study. The patients included 26 men and 10 women with a mean age of 60 years (range 43-79 years). All patients were imaged on a 16-row multidetector CT scanner. Mean value of BW and BMI was 83.5 kg and 28.1, respectively. Image noise was defined as standard deviation (SD) of the attenuation values measured by using 1 cm2 circular region of interest in the ascending aorta at the level of the right main pulmonary artery. The SD values were plotted against BW and BMI. The correlations were examined using a linear regression method. A P value of less than .05 was considered significant. RESULTS: The r value of linear regression between noise and BW was 0.90 (P < .001). The r value of linear regression between noise and BMI was 0.74 (P = .015). CONCLUSIONS: Excellent correlation was observed between noise and BW in coronary CTA. These data may be used as potential parameters for customized radiation dose modification to reduce radiation dose in coronary CT examinations.

Adult↗
PubMed2003

[Selecting methods of controls concentration for internal quality control and continuity of control chart between different reagent lots for HBsAg qualitative detection].

OBJECTIVE: To establish a model for one choosing controls with a suitable concentration for internal quality control (IQC) with qualitative ELISA detection, and a consecutive plotting method on Levey-Jennings control chart when reagent kit lot is changed. METHODS: First, a series of control serum with 0.2, 0.5, 1.0, 2.0 and 5.0ng/ml HBsAg respectively were assessed for within-run and between-run precision according to NCCLs EP5 document. Then, a linear regression equation (y=bx + a) with best correlation coefficient (r > 0.99) was established based on S/CO values of the series of control serum. Finally, one could choose controls with S/CO value calculated from the equation (y = bx + a) minus the product of the S/CO value multiplying three-fold between-run CV to be still more than 1.0 for IQC use. For consecutive plotting on Levey-Jennings control chart when ELISA kit lot was changed, the new lot kits were used to detect the same series of HBsAg control serum as above. Then, a new linear regression equation (y2 = b2x2 + a2) with best correlation coefficient was obtained. The old one (y1 =b1x1 + a1) could be obtained based on the mean values from above precision assessment. The S/CO value of a control serum detected by new lot kit could be changed to that detected by old kit lot based on the factor of y2/y1. Therefore, the plotting on primary Levey-Jennings control chart could be continued. RESULTS: The within-run coefficient of variation CV of the ELISA method for control serum with 0.2, 0.5, 1.0, 2.0 and 5.0ng/ml HBsAg were 11.08%, 9.49%, 9.83%, 9.18% and 7.25%, respectively, and between-run CV were 13.25%, 14.03%, 15.11%, 13.29% and 9.92%. The linear regression equation with best correlation coefficient from a test at random was y = 3.509x + 0.180. The suitable concentration of control serum for IQC could be 0.5ng/ml or 1.0ng/ml. The linear regression equation from the old lot and other two new lots of the ELISA kits were y1 = 3.550(x1) + 0.226, y2 = 3.238(x2) +0.388, and y3 =3.428(x3) + 0.148, respectively. Then, the transferring factors of 0.960 (y2/y1) and 0.908 (y3/y1) were obtained. CONCLUSION: The results shows that the model established for IQC control serum concentration selecting and for consecutive plotting on control chart when the reagent lot is changed is effective and practical.

Enzyme-Linked Immunosorbent Assay↗
PubMed2003

Indirect method for prediction of hemagglutination inhibition antibody titers to Newcastle disease virus in chickens by titration of antibodies in egg yolk.

Attempts were made to establish methods for indirect prediction of hemagglutination inhibition (HI) antibody titers to Newcastle disease virus (NDV) in sera of laying hens and day-old chicks by determining if these are correlated to HI titers in egg yolks. For this purpose, geometric means of HI antibody titers in sera from 60 hens, yolks from 60 matched eggs, and sera from 180 day-old chicks of an identical vaccination program were measured and plotted. There was a significant correlation between HI antibody titers in yolks (X) and hens (Y), with a linear regression of Y = 23.24 + 0.47X and a correlation coefficient of r = 0.65. The linear regression between HI antibody titers in yolks (X) and chicks (Y) was Y = 6.33 + 0.36X (r = 0.58). Immunity to NDV in hens and their offspring can be maintained effectively, and the proper time for the vaccination or booster can be determined by reference to HI titers predicted from the linear regression in the present study. The approach of testing egg yolk for HI titers provides a feasible alternative to determining HI titers from blood samples and eliminates stress in birds during blood sampling.

Animals↗
PubMedDate not supplied

A computerized method for measuring respiratory mechanics during mechanical ventilation.

In order to monitor respiratory mechanics in anesthetized ventilated subjects, an automated method was developed and tested on a physical model. The model was composed of an endotracheal tube (curvilinear resistance) coupled to a rigid air-filled box (elastance). Theoretical resistance was determined during steady-state flow experiments and theoretical elastance was estimated from the dimensions of the box. The physical model was connected to a volume-cycled ventilator. Pressure (P) and air flow (V) were measured at the outlet of the tube, and the time integral V of V was calculated. Elastance (E) and curvilinear resistance (R + K/V/) were identified by multiple linear regression analysis, for each ventilatory cycle, according to equation P = EV + RV + K/V/V. When linear regression analysis of P on V, V and /V/V was performed over the entire ventilatory cycle, E was found equal to its estimate, whereas R and K appeared different from their theoretical values. In order to improve determination of R and K, resistive pressure (Pr = P - EV) was calculated using the previously obtained value of E, and multiple linear regression of Pr on V and /V/V was performed over different fractions of the ventilatory cycle. When determined over the expiratory phase corresponding to decreasing flows, R and K were found close to their expected values. Such a method to calculate elastance and curvilinear resistance should prove convenient and efficient in measuring respiratory mechanics during mechanical ventilation.

Computers↗
PubMed1988

Hierarchical time-oriented approaches to missing data inference.

In practice clinical data are nearly always incomplete. When confronted with such data, a physician or investigator must make inferences about missing information. Possible strategies for inference include (1) interpolation, (2) extrapolation, (3) repeating the nearest value, (4) repeating the previous value, (5) patient-specific mean values, (6) patient-specific linear regression over time, (7) disease-specific mean values, (8) normal values, and (9) linear regression of correlated co-recorded variables. This study analyzes these strategies in a time-oriented data bank of patients with systemic lupus erythematosus, demonstrating that more accurate inferences of missing data are obtained when (1) strategies are tailored to the characteristics of the individual variable, (2) time-oriented strategies (e.g., interpolation) rather than non-time-oriented strategies (e.g., disease mean) are incorporated, (3) a ranked set of strategies is incorporated in a hierarchical stepwise fashion, and (4) the degree to which missing data are "nonrandomly" missing is assessed to allow estimation of bias. Interpolation is the best single technique with these data while linear regression of correlated co-recorded variables is a relatively weak technique. Inferences made by these hierarchical time-oriented approaches show significantly smaller mean differences from the actual values than do results from typical statistical package strategies.

Data Interpretation, Statistical↗
PubMed1999

An application of lifetime models in estimation of expected length of stay of patients in hospital with complexity and age adjustment.

Expected length of stay (ELOS) of patients in hospital is an important measure in hospital resource utilization management. Previous work has shown that estimation of ELOS is improved using complexity and age adjustment. These improved estimates have the potential to improve the accuracy of estimates of resource use. Recently other authors have applied the linear regression model to make complexity and age adjustments in the estimation of ELOS. However, these estimates using linear regression estimates are likely flawed on the basis that the assumptions regarding the distribution of data for the linear regression model are unjustifiable. The non-normal distributions of most hospital patient discharge data demand that an alternative method be described to provide accurate estimates of ELOS. The purpose of this paper is to describe an alternative method which uses lifetime models to initially estimate the expected length of stay. The paper then provides an approach to estimate the adjusted expected length of stay (AELOS) using several influencing factors by application of lifetime models. Depending on whether or not the proportional hazards assumption is appropriate for the data, the Cox proportional hazards model or the Kaplan-Meier adjustment is recommended.

Age Factors↗
PubMed1990

Practical means for estimating pork carcass composition.

Three hundred sixty-one market-weight barrow and gilt carcasses were physically dissected into bone, skin, fat and muscle. A three-variable multiple linear regression equation containing the same independent variables (warm carcass weight, 10th rib loin muscle area and 10th rib fat depth) used (U.S.) to determine pork carcass lean weight was found to be the most practical means for predicting weight of muscle standardized to 10% fat. Multiple linear regression equations containing more than three independent variables produced only slight improvements in R2 values; however, the standard deviation about the regression line was not greatly improved by the addition of more independent variables to this three-independent-variable regression model. A single multiple linear regression equation using the three independent variables above may not be adequate to describe variation over the entire live-weight range for all hogs marketed in the U.S. For most accurate muscle weight prediction, different equations should be used for weight subclasses with one equation for carcasses under 100 kg and another for those heavier than 100 kg. A single prediction equation for muscle weight was adequate for carcasses of both barrows and gilts.

Adipose Tissue↗
PubMed2001

Rapid assessment of the bacteriological quality of raw milk using ATP bioluminescence.

Research was conducted to assess the practical use of an ATP bioluminescence assay to evaluate the bacteriological quality of raw milk. Filtration was used to precondition samples before ATP determination, which was measured in relative light units (RLUs). The Lumac ATP bioluminescence assay results were compared with standard plate counts (SPCs) of samples to estimate the microbial load for 246 raw milk samples that were split and either tested immediately or subjected to two preliminary incubation temperatures, 12.8 and 15.6 degrees C, for 18 h. Linear regression analysis procedures were used to analyze the data. Preincubation treatments were analyzed separately. For all treatments, linear regression coefficients were significantly different from zero (P < 0.01). The R2 values calculated using log10-transformed SPC and log10-transformed RLUs for samples tested immediately and samples preliminarily incubated at 12.8 and 15.6 degrees C were 0.58, 0.78, and 0.80, respectively. The R2 for all samples combined was 0.78. Differences in regressions among treatments were tested using a multiple slope and intercept model. Treatment intercepts and slopes were significantly different (P < 0.01). A linear regression equation was used to predict SPC from ATP values. Comparison of predicted values with actual SPCs indicated that ATP could be useful in predicting SPC in raw milk.

Adenosine Triphosphate↗
PubMed2002

Sensitization of glutamate release and N-methyl-D-aspartate receptor response by transient dopamine pretreatment in prefrontal cortex of rats.

We studied biochemically the effect of transient dopamine pretreatment on the regulation of glutamate transmission in medial prefrontal cortex of rats in vivo and in vitro. Aversive stimuli transiently increased the glutamate concentration and its repetition reduced the response in the medial prefrontal microdialysate of freely moving rats. The rate of habituation obeyed linear regression. The medial prefrontal intracellular calcium response to repetitive N-methyl-D-aspartate perfusion showed linearly regressive desensitization in fluorescence videomicroscopy of the fura-2 stained slice in vitro. Transient dopamine treatment 10-20 min prior to repetition restored both decreased responses in a linearly regressive manner, also indicating that their decrease was not due to fatigue. These findings suggest that the effect of transient dopamine pretreatment continues redundantly to sensitize/resensitize subsequent pre- and postsynaptic prefrontal glutamate transmission in an orderly manner.

Animals↗
PubMed2002

Predicting survival of a genetically engineered microorganism, Pseudomonas chlororaphis 3732RN-L11, in soil and wheat rhizosphere across Canada with linear multiple regression models.

Pseudomonas chlororaphis 3732RN-L11 survival rates in soil and wheat rhizosphere were measured using intact soil core microcosms representing 23 sites across Canada. Linear multiple regression (LMR) models were developed to predict the survival rate of this genetically engineered microorganism (GEM) as a function of soil parameters measured at the time of microcosm inoculation. LMR models were tested by comparing their predicted survival rates with observed survival rates from environmental introductions of the GEM by Gagliardi et al. (2001) at five field sites across Canada over two years. No soil parameter (e.g., % clay) was highly correlated with GEM survival rates in soil or wheat rhizosphere. Total fungal colony-forming units (CFUs), % soil titanium (positive correlations), and % soil magnesium (negative correlation) were found to be the best LMR predictors of GEM survival rates in soil over two years. Total soil bacterial CFUs, nitrate, % soil potassium (positive correlations), and exchangeable magnesium (negative correlation) were found to be the best LMR predictors of GEM survival rate in wheat rhizosphere over two years. While LMR models were statistically significant, they were unable to reliably predict the survival rate of the GEM in field trial introductions. The results indicate that there can be considerable uncertainty associated with predicting GEM survival for multi-site environmental introductions.

Canada↗