Open AccessModeling of Sea Surface Temperature through Fitting Linear Model with Interaction
Author(s) -
Miftahuddin Miftahuddin,
Wanda Sri Noviana
Publication year - 2021
Publication title -
jurnal matematika, statistika dan komputasi/jurnal matematika statistik dan komputasi
Language(s) - English
Resource type - Journals
eISSN - 2614-8811
pISSN - 1858-1382
DOI - 10.20956/j.v18i1.13987
Subject(s) - covariate , autoregressive model , autocorrelation , statistics , sea surface temperature , mathematics , linear model , transformation (genetics) , linear regression , econometrics , climatology , biochemistry , chemistry , gene , geology
Sea surface temperature (SST) is one of the attributes of the world climate system and global warming. The relationship between SST and other climate parameters can be represented in a linearity approach. Through this approach, SST variability shows monthly and yearly effects. Information on these two time effects is important for knowing the period of peak effect as well as other statistical measures in the linear fitting model. The models used include transformation and without covariate transformation, interaction and without covariate interaction, and with centering and with the addition of time covariates in the model. The linear fitting model chosen as the basis for construction is a model with a combination effect of covariate interaction and transformation giving an increase in the magnitude of multiple R2 (56.62%) and adjusted R2 (56.13%) respectively 0.31% and 0.43%. This indicates that the time covariate has a very strong significant effect on the model compared to the continuous covariate. In general, the model has a statistical significance of p-value < 2.2e-16, as well as for the time covariate. However, because the model has an autocorrelation and a large AIC value, this effect is removed by means of an autoregressive moving average. The obtained linear fitting model for SST data is the model with AIC 403.2987.