Page 46 - ISMCON souvenir 2021

ISMSCON - 2021

(15-49 years). Methods: It was a cross sectional study done on women in the age group of 15 – 49
years residing in rural field practice area of NMCH, Sasaram and willing to participate in the study
excluding pregnant women. Simple random sampling was used for the selection of study participants
which were 283 in number. Duration of the study was six months. Templates were generated in MS
Excel sheet and analysis was done using SPSS software version 20. Results: After analysis it was
found that history of blackouts, cycle regularity was significantly associated with anaemia. Other factors
associated significantly were handling money matters, participation in decision making and relationship
with family members (p<0.05). Conclusions: In this COVID-19 pandemic many essential services are
being affected. It was concluded from the study that anaemia being a public health problem specially in
women of reproductive age group (15-49 years) is affected by various factors.
Keywords: Anaemia, Women of Reproductive Age Group (15-49 years), COVID-19 Pandemic

OS3: APPLICATION OF MULTIVARIATE APPROACH TO ANALYZE

MULTIPLE CORRELATED OUTCOMES AND CONTRIBUTION
OF EACH VARIABLE TOWARDS REJECTON OF MULTIVARIATE

HYPOTHESIS

Akash Mishra1, N. Sreekumaran Nair2, KT.Harichandrakumar3, Binu VS.4, Santhosh Satheesh5
Author affiliation:
1 Ph.D. Scholar, Department of Biostatistics, Jawaharlal Institute of Postgraduate Medical Education and
Research, Puducherry, India;
Email addresses: akashmishra0292@gmail.com

ABSTRACT
Background: In clinical research, the hypothesis testing for two independent group comparison often
based on multiple outcome variables which could be correlated. The usual analysis procedure is univariate
approach where each variable is analyzed separately by treating them as independent which ends up
with biased decision. The multivariate approach which captures the correlation between the variables
could give more robust decision. The objective of this paper is to demonstrate the change in statistical
decision in multivariate approach compared to univariate approach and further to see the contribution of
each variable towards the rejection of multivariate hypothesis.
Methodology: The ACCORD Lipid trial dataset was used for demonstration. The correlated lipid variables
chosen were TG, HDL, LDL at baseline, 12th and 36th month. The condition of multivariate normalcy was
checked after removing the outliers. The student independent t test was used in univariate approach and
Hotelling’s T2 in multivariate approach. Further, the contribution of each variable towards the rejection of
multivariate hypothesis was demonstrated by using standard discriminant function coefficient and partial
F test.
Results: At baseline, the univariate as well multivariate approach showed two groups are statistically
similar. At 12th and 36th months in univariate analysis TG and HDL found to differ significantly whereas
LDL was insignificant. The multivariate approach rejected the hypothesis at these two follow ups. The
TG contributed the most at 12th and 36th month followed by HDL and then LDL towards the rejection of
multivariate hypothesis.
Conclusion: The studies with multiple correlated outcomes should be analyzed using multivariate
approach for more valid decision. Further, the relative importance of each variable could also be known
for rejection towards multivariate hypothesis.
Keywords: multiple correlated outcomes, Multivariate approach, Hotelling’s T2, Standardized
discriminating function coefficient, Partial F test.

44 CONFERENCE SOUVENIR