Robustness in Regression Analysis

Bohrnstedt, George W.; Carter, T. Michael

doi:10.2307/270820

Cited by 192 publications

(87 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…However, despite being a sophisticated data analysis technique, multiple regression is not robust to measurement error and usually assumes perfect measurement of variables. Yet, perfect reliability is rarely achieved in social science research (Bohrnstedt & Carter, 1971;Musil et al, 1998). Measurement error in even a single independent variable can lead to statistical inaccuracies and biased results (Mackenzie, 2001;Musil et al, 1998;Nusair & Hua, 2010) by increasing or decreasing all other regression coefficients in a multiple regression equation.…”

Section: Multiple Regressionmentioning

confidence: 99%

Place attachment, place satisfaction and pro-environmental behaviour: a comparative assessment of multiple regression and structural equation modelling

Ramkissoon

Weiler

Smith

2013

Journal of Policy Research in Tourism, Leisure and Events

View full text Add to dashboard Cite

Using data collected from 452 visitors at the Dandenong Ranges National Park, Australia, this paper applies multiple regression and structural equation modelling (SEM) approaches to test the relationships between the respective sub-constructs of place attachment (place dependence, place identity, place affect, place social bonding), place satisfaction, and visitors' pro-environmental behavioural intentions. Findings revealed differences in strength and magnitude of the regression paths. The SEM analysis further reveals insignificant regression paths between place identity and place satisfaction, and place social bonding and low effort pro-environmental behavioural intentions in contrast to the multiple regression results. An empirical comparison of the MR and SEM results is expected to shed light on the application of these two powerful techniques in tourism research.

show abstract

Section: Multiple Regressionmentioning

confidence: 99%

Place attachment, place satisfaction and pro-environmental behaviour: a comparative assessment of multiple regression and structural equation modelling

Ramkissoon

Weiler

Smith

2013

Journal of Policy Research in Tourism, Leisure and Events

View full text Add to dashboard Cite

show abstract

“…Labovitz 1967Labovitz , 1970Bohrnstedt & Carter, 1971;Asher, 1976;Nunnally, 1978). 7 When variance inflation factors (VIP) were examined for predictors, no concern was found for multicollinearity: the simple correlations were quite low and all VIPs were less than 1.263 ( i.e., substantially below 10, see Stevens, 2002).…”

Section: Notesmentioning

confidence: 99%

The determinants of multiple forms of controls in foreign subsidiary manager evaluations

Kihn

2008

IJAAPE

View full text Add to dashboard Cite

“…Some might argue on theoretical grounds that HLM might offer a better overall solution (''fit'') to the multivariate data, but this is by no means a given. Over the years, OLS regression has proven to be a very robust method, one that produces a reasonably good fit even when some of its assumptions (e.g., homoscedasticity, linearity) are not satisfied (e.g., Bohrnstedt and Carter 1971;Hanushek and Jackson 1977;Snedecor and Cochran 1967). Moreover, on a purely practical level, the existing HLM 6.0 software package (Raudenbush et al 2004) does not offer the same options as the SPSS software package (e.g., blocking, forward entry, ''Beta in'') that make it possible to conduct the kinds of modeling of variables described in the discussion of OLS regression.…”

Section: Hlmmentioning

confidence: 99%

Multi-Campus Studies of College Impact: Which Statistical Method is Appropriate?

Astin

Denson

2009

Res High Educ

View full text Add to dashboard Cite

In most multi-campus studies of college impact that have been conducted over the past four decades, investigators have relied on ordinary least squares (OLS) regression as the analytic method of choice. Recently, however, some investigators have advocated the use of Hierarchical Linear Modeling (HLM), a method specifically designed for analyses that involve both individual (student) and aggregate (institutional) level measures. Cross-validation analyses using a national database show that the two methods yield an equally good ''fit'' with empirical data. Existing OLS software has the advantage of enabling one to perform path analytical causal modeling; HLM has the advantage of yielding a more conservative estimate of the significance of institution-level effects.Keywords Ordinary least squares Á Hierarchical linear modeling Á Methodology Á College effects Á Stepwise regression Studies of student development that involve longitudinal data from several campuses have one obvious advantage over single-campus studies: they enable the researcher to obtain empirical estimates of (1) the comparative effects of different institutions and (2) the effects of institutional characteristics on students. These characteristics typically include measures such as size and selectivity, but-depending on the data available to the researcher-they can also include more subtle attributes such as peer group measures and characteristics of the faculty.

show abstract

Robustness in Regression Analysis

Cited by 192 publications

References 34 publications

Place attachment, place satisfaction and pro-environmental behaviour: a comparative assessment of multiple regression and structural equation modelling

Place attachment, place satisfaction and pro-environmental behaviour: a comparative assessment of multiple regression and structural equation modelling

The determinants of multiple forms of controls in foreign subsidiary manager evaluations

Multi-Campus Studies of College Impact: Which Statistical Method is Appropriate?

Contact Info

Product

Resources

About