Statistics notes: Transformations, means, and confidence intervals

Bland, John M.; Altman, D G

doi:10.1136/bmj.312.7038.1079

Cited by 370 publications

(271 citation statements)

References 0 publications

Supporting

Mentioning

267

Contrasting

Unclassified

Order By: Relevance

“…Correlations were obtained by linear regression. Triglyceride data were log 10 transformed before statistical analyses [27]. Analyses were made using SPSS and the level of statistical significance was set to p≤0.05.…”

Section: Statistical Methodologymentioning

confidence: 99%

The effect of sensitisation to insulin with pioglitazone on fasting and postprandial lipid metabolism, lipoprotein modification by lipases, and lipid transfer activities in type 2 diabetic patients

et al. 2006

View full text Add to dashboard Cite

Aims/hypothesis: Insulin resistance is thought to be central to the pathogenesis of diabetic dyslipidaemia. We hypothesised that improving insulin sensitivity would improve fasting and postprandial triglyceride metabolism in patients with type 2 diabetes. To this aim we studied fasting and postprandial lipaemia in type 2 diabetic patients before and after sensitisation to insulin with pioglitazone, compared with that observed in patients on an insulinproviding regime. Methods: In a double-blind placebo-controlled protocol, 22 patients with type 2 diabetes were randomly allocated to receive either pioglitazone (45 mg/day) or glibenclamide (5 mg/day), for a 20-week period. Fasting and postprandial lipid metabolism were investigated at baseline and at the end of the treatment period. A group of non-diabetic subjects was also studied. Results: Compared with glibenclamide treatment, pioglitazone treatment decreased fasting triglyceride, glucose and insulin levels and the homeostasis model assessment score of insulin resistance. Decreased fasting triglyceride after pioglitazone treatment was due to reduced VLDL triglyceride, particularly VLDL-2. Lipoprotein lipase activity was unchanged by pioglitazone treatment but hepatic lipase showed a significant decrease. Pioglitazone treatment lowered total postprandial triglyceride, as well as chylomicron-and chylomicron-remnant retinyl palmitate levels to normal. Glucose disposal improved but remained abnormal. Conclusions/interpretation: Insulin sensitisation with pioglitazone has major effects in restoring postprandial lipaemia to normal, while also correcting fasting hypertriglyceridaemia; both factors may have consequences for atherogenic risk in diabetes.

show abstract

Section: Statistical Methodologymentioning

confidence: 99%

The effect of sensitisation to insulin with pioglitazone on fasting and postprandial lipid metabolism, lipoprotein modification by lipases, and lipid transfer activities in type 2 diabetic patients

et al. 2006

View full text Add to dashboard Cite

show abstract

“…Finally, data were transformed so that the ordering of values was identical to that in the raw data with the addition of 1 to scores to allow calculation of geometric means and standard deviations and to minimize the impact of extreme responses (eg, ''1'' and ''2'') on estimates of central tendency. 16 Linear trend tests (polynomial contrast) were conducted under the generalized estimating equations model to assess the relation between the number of tests requested and willingness to undergo the tests. The main effect of test type also was assessed to determine whether participants differed, on average, in their willingness to undergo different tests.…”

Section: Discussionmentioning

confidence: 99%

Patient willingness to undergo pharmacodynamic and pharmacokinetic tests in early phase oncology trials

Tibes

Piper

Smith

et al. 2011

Cancer

View full text Add to dashboard Cite

BACKGROUND: Increasingly, early phase clinical trials involve pharmacodynamic (PD) and pharmacokinetic (PK) assays as well as frequent imaging studies. The authors conducted a prospective study examining patients' willingness to undergo such tests and the number of tests the patients would tolerate. METHODS: A prospective, correlative study was conducted using a self-reported questionnaire to measure patients' willingness on a scale from 1 (not willing) to 10 (very willing) to undergo various procedures (eg, tumor and skin biopsies, blood tests) and imaging studies (eg, magnetic resonance imaging, echocardiogram). In addition, correlations were assessed between the number and type of tests and demographics, previous test experience, inconvenience, and insurance coverage. Sixty-one patients (22 women and 39 men) with advanced malignancies were enrolled. Descriptive, nonparametric, and parametric inferential statistics were used. RESULTS: Overall willingness to undergo study-required tests was very high. Patients were most willing to undergo urine, blood, ultrasound, x-rays, echocardiogram, and computed tomography studies and were least willing to undergo tumor and skin biopsies and magnetic resonance imaging (all P .01). Significant inverse relations were observed between the frequency of a particular test and patient's willingness to undergo such tests. Inconvenience and prior negative experiences for more invasive tests (eg, skin biopsies) modestly affected willingness to undergo these tests again. College education, insurance coverage, and the requirement of tests for enrollment were correlated positively with willingness to undergo tests. CONCLUSIONS: The current findings provide the first prospectively collected data on patients' willingness to undergo PK/PD tests and imaging studies associated with early stage oncology drug trials and can serve as basis for further exploration toward the design of patientfriendly, biomarker-driven clinical studies in oncology. Cancer 2011;117:3276-

show abstract

“…Instead of always using a log transformation, a more robust approach is to select a transformation that best normalizes the data. While this can be accomplished using the Box-Cox power transformation that automatically selects the best transformation (Box & Cox, 1964), it still fails to normalize the data in some cases. However, transformations are not effective when the data have an extremely large accumulation of zeros, as in the case of pressure ulcers.…”

Section: Resultsmentioning

confidence: 99%

“…Similarly, if the upper limit exceeded 100%, it would be truncated at 100%. For the transformation method, we used a log 10 transformation that is a popular choice of statisticians to help normalize data that are positively skewed (Bland & Altman, 1996). This is an appropriate choice because the indicators that are skewed typically peak around low values and tail off to larger values.…”

Section: Methodsmentioning

confidence: 99%

Summarizing benchmarks in the national database of nursing quality indicators using bootstrap confidence intervals

Gajewski

Hall

Dunton

2007

Research in Nursing & Health

View full text Add to dashboard Cite

When summarizing the benchmarks for nursing quality indicators with confidence intervals around the means, bounds too high or too low are sometimes found due to small sample size or violation of the normality assumption. Transforming the data or truncating the confidence intervals at realistic values can solve the problem of out of range values. However, truncation does not improve upon the non-normality of the data, and transformations are not always successful in normalizing the data. The percentile bootstrap has the advantage of providing realistic bounds while not relying upon the assumption of normality and may provide a convenient way of obtaining appropriate confidence intervals around the mean for nursing quality indicators. ß

show abstract

Statistics notes: Transformations, means, and confidence intervals

Cited by 370 publications

References 0 publications

The effect of sensitisation to insulin with pioglitazone on fasting and postprandial lipid metabolism, lipoprotein modification by lipases, and lipid transfer activities in type 2 diabetic patients

The effect of sensitisation to insulin with pioglitazone on fasting and postprandial lipid metabolism, lipoprotein modification by lipases, and lipid transfer activities in type 2 diabetic patients

Patient willingness to undergo pharmacodynamic and pharmacokinetic tests in early phase oncology trials

Summarizing benchmarks in the national database of nursing quality indicators using bootstrap confidence intervals

Contact Info

Product

Resources

About