The fisheries sector has an important role in supporting the food security chain, where the world's protein needs can be met by fisheries resources, both from capture fisheries and aquaculture. There are several fisheries sectors including fishing companies, capture fisheries production, number of ships, types and size of cultivated land. Therefore a statistical analysis is needed to increase the potential of fisheries in Indonesia. Data on the fisheries sector used in this study from the Indonesian Central Statistics Agency in 2018, which included the 2016 fisheries sector with 34 observation units in Indonesia. By using cluster analysis K-Means aims to group provinces in Indonesia based on the fisheries sector so that several groups are formed which will show the characteristics of each group. There are three determinations of the optimum number of clusters, namely the Elbow method, Silhouette method, and GAP Statistics. The results showed that optimum clusters were formed in 2 clusters, with the best Elbow and Silhouette methods. Where the first cluster is a region that shows a low value of the fisheries sector consisting of 30 provinces this is due to inadequate infrastructure and use that is not optimal while cluster 2 regions that have great potential in the Indonesian fisheries sector in 2016 as many as 4 provinces namely West Java, Java Central, East Java, and South Sulawesi as dominating capture fisheries production and aquaculture. Keywords: Fisheries Sector, K-Means Cluster Analysis, Elbow Method, Silhoutte Method and GAP Statistics.

ABSTRAK Dalam analisis regresi, salah satu asumsi yang harus dipenuhi adalah tidak adanya hubungan antar variabel independen. Hubungan yang kuat antar variabel independen disebut dengan multikolinieritas. Berbagai metode dapat menanggulangi kasus multikolinieritas, semua itu bergantung pada tujuan dari penelitian. Beberapa metode tersebut adalah ridge regression, principal component regression, regresi robust dan pemilihan model terbaik. Pada penelitian ini, metode pemilihan model terbaik dipilih untuk digunakan karena bertujuan untuk menentukan variabel independen yang signifikan dengan mempertimbangkan korelasi parsial pada data track quality index (TQI) kereta api Indonesia. Untuk mengukur besarnya TQI diperlukan empat indikator yang kemudian menjadi variabel dalam penelitian ini, yaitu lebar jalur, angkatan, listringan dan pertinggian. Hasil analisis menunjukkan variabel pertinggian, angkatan dan listringan berpengaruh besarnya nilai TQI dengan variasi data yang dapat dijelaskan model sebesar 99,7%.Kata kunci: multikolinieritas, stepwise, track quality index. ABSTRACTIn regression analysis, one of the assumptions is the absence of relationships between independent variables. Relationship between independent variables is called multicollinearity.Various methods can overcome multicollinearity cases, all of which depend on the purpose of the study. Some of these methods are ridge regression, principal component regression, robust regression and selection of the best models. In this study, the best model selection method was chosen because it aims to determine significant independent variables taking into account the partial correlation of track quality index (TQI) data. To measure the magnitude of TQI, four indicators are needed which then become the variables in this study, namely the width of the track, force, lightning and elevation. The results of the analysis show that the height, force and list

No abstract

Kebijakan system pembelajaran daring dalam Rangka Pencegahan Penyebaran COVID-19 menuntut mahasiswa untuk segera beradaptasi dengan berbagai macam bantuan teknologi yang telah berkembang cukup pesat di era saat ini. Interaksi mahasiswa dan dosen menjadi berkurang, begitu juga antar mahasiswa yang semakin susah untuk saling berdiskusi. Interaksi dan perilaku sosial yang efektif akan memberikan dampak terhadap proses pembelajaran yang efektif pula. Sehingga untuk mengetahui pengaruh perubahan perilaku sosial mahasiswa terhadap efektifitas pembelajaran di tengah pandemi, perlu dilakukan kajian guna menciptakan sistem pembelajaran yang lebih efektif, nyaman, dan efisien. Salah satu metode yang dapat digunakan untuk mengetahui pengaruh tersebut yaitu dengan analisis regresi logistik biner. Analisis regresi logistik biner merupakan salah satu metode yang dapat menggambarkan hubungan antara variabel respon (berskala biner yaitu mempunyai dua kategori) dan variabel independennya. Hasil penelitian diperoleh bahwa 85% dari mahasiswa yang dijadikan sebagai sampel penelitian merasa bahwa pembelajaran daring sudah efektif pelaksanaannya. Diperoleh variabel yang berpengaruh signifikan terhadap efektifitas pembelajaran adalah variabel Interaksi dengan Mahasiswa dan variabel Perilaku Belajar. Hasil penelitian menunjukkan bahwa mahasiswa akan berpeluang mengalami pembelajaran yang efektif sebesar 99,98% apabila mengalami perubahan positif terhadap interaksi antar mahasiswa dan perubahan positif terhadap perilaku belajarnya selama pembelajaran daring.

Monte Carlo is a method used to generate data according to the distribution and resampling until the parameters of the method used became convergen. The purpose of this simulation is first to prove that quantile regression with the estimated sparsity function parameter can model the data according to the non-uniform distribution of the data. Secondly, it’s to prove that the quantile regression is a developed method from the linear regression. The pattern of data which is not uniform is generally referred to as heterogeneous data, while the pattern of uniform data distribution is called homogeneous data. Data in this study will be generated for small and large samples on homogeneous and heterogeneous data. Uniformity of variance will be carried out on both heterogeneous and homogeneous data types, namely 0.25,1 and 4. The parameter estimation process and data generation will be resampled 1000 times. Thus, in conclusion of the simulation studies was the parameter estimates in the classical regression will be the same as the parameter estimates in the quantile regression at quantile 0.5. In the simulation, it is decided that the quantile regression method can be used on heterogeneous and homogeneous data to the unconstrained number of samples and variances.