Peran Pengukuran Dananalisis Statistika Dalam Penelitian Psikologi

Jahja Umar (Institut Asesmen Indonesia)

Jurnal Pengukuran Psikologi dan Pendidikan Indonesia (JP3I)
Vol 4, No 1 (2015): JP3I

Abstract At least, there are four source of error which cause research result invalidity, especially in data analysis using statistics, those are sampling error, measurement error, specification error, other sources of error that haven't been know and assumed as random error. In psychology research, impact of measurement error to research result validity more dominantly is caused by almost all variables in psychology that are latent. Impact of measurement error generally is in form of under-estimation to statistical coefficient such as correlation, regression, etc. But now, there are methods that can be used to correct all those negative impacts, such as confirmatory factor analysis which result true-score estimation to be used as data that will be analysed. There is also structural equation model analysis where error impact can corrected directly, thus pure correlation and regression coefficient obtained. This article illustrates how serious error that can happen without correction, using SEM approach   Abstrak Setidaknya ada empat sumber kesalahan yang dapat mengakibatkan tidak validnya hasil penelitian,terutamadalamanalisisdatastatistika,yaitu:(a)kesalahansampling,(b)kesalahanpengukuran,(c)kesalahanspesifikasi/teori ,dan(d)sumberkesalahanlainyangbelumdiketahuidandianggapbersifat random.Dalampenelitianpsikologidampak kesalahan pengukuran terhadap validitas hasil penelitian lebih dominankarenahampirsemuavariabelnya merupakan variabel laten. Dampak dari kesalahan pengukuran umumnyaberupaunder-estimasi terhadap koefisien statistik seperti korelasi, regresi, dan lain-lainnya.Namunpadasaatini,telah tersedia berbagai metode statistika yang dapat digunakan untuk mengoreksi dampak negatif tersebut, misalnya analisis faktor konfirmatorik yang dapatmenghasilkanestimate true-scoreuntukdigunakansebagaidatayang akandianalisis.Selanjutnyajugaterdapatanalisismodelpersamaanstrukturaldimana dampak kesalahan pengukuran dapat dikoreksi secara langsung sehingga diperoleh koefisien korelasidanregresiyangmurni.Dalamtulisaninidisajikan ilustrasi (dengan pendekatan SEM) betapa seriusnya kesalahan yang dapat terjadi jika tidak dilakukankoreksi.