مقایسه روشهای مختلف برآورد واریانس با استفاده از مدلهای مختلط در تجزیه پایداری
پذیرفته شده برای پوستر
عنوان دوره: سیزدهمین کنگره زراعت و اصلاح نباتات ایران
نویسندگان
1دانش آموخته گروه اصلاح نباتات، دانشکده کشاورزی، دانشگاه تربیت مدرس
2دانشیار گروه اصلاح نباتات، دانشکده کشاورزی، دانشگاه تربیت مدرس
3مربی تحقیقاتی، ایستگاه تحقیقات کشاورزی دیم گچساران
چکیده
به ﻣﻨﻈﻮر ﻣﻘﺎﻳﺴﻪ روشﻫﺎی ﻣﺨﺘﻠﻒ ﺑﺮآورد وارﻳﺎﻧﺲ ﺑﺎ اﺳﺘﻔﺎده از ﻣﺪلﻫﺎی ﻣﺨﺘﻠﻂ، ﻣﻄﺎﻟﻌﻪای ﻃﻲ ﭼﻬﺎر ﺳﺎل زراﻋﻲ در اﻳﺴﺘﮕﺎهﻫﺎی ﺗﺤﻘﻴﻘﺎﺗﻲ دﻳﻢ ﮔﭽﺴﺎران، ﻛﻮﻫﺪﺷﺖ و اﻳﻼم اﺟﺮا ﺷﺪ. ﺗﺠﺰﻳﻪ وارﻳﺎﻧﺲ ﻣﺮﻛﺐ ﺑﺎ اﺳﺘﻔﺎده ازﻣﺪل ﻣﺨﺘﻠﻂ اﻧﺠﺎم ﺷﺪ. ﺳﭙﺲ اﺛﺮات وارﻳﺎﻧﺲﻫﺎ ﺑﺮ اﺳﺎس روش TYPEIII، ﺣﺪاﻛﺜﺮ درﺳﺖﻧﻤﺎﻳﻲ )ML( و ﺣﺪاﻛﺜﺮ درﺳﺖﻧﻤﺎﻳﻲ ﻣﺤﺪودﺷﺪه(REML) ﺑﺎ اﺳﺘﻔﺎده از روﻳﻪ ﻣﺨﺘﻠﻂ ﺑﻪدﺳﺖ آﻣﺪﻧﺪ. ﻧﺘﺎﻳﺞ ﻧﺸﺎن داد ﻛﻪ اﺛﺮ ﻣﻜﺎن در ﻃﻲ 4 ﺳﺎل ﺑﺎ ﻫﺮ ﺳﻪ روش ﻣﺬﻛﻮر ، ﻣﻌﻨﻲدار ﻧﺸﺪ. ﻫﻤﭽﻨﻴﻦ اﺛﺮ ﻣﺘﻘﺎﺑﻞ ژﻧﻮﺗﻴﭗ ×
ﻣﻜﺎن ﺗﻨﻬﺎ در ﺳﺎل دوم P<0.01( ﻣﻌﻨﻲدار ﺷﺪ. ﻫﻤﭽﻨﻴﻦ اﺛﺮ ژﻧﻮﺗﻴﭗ، ﺗﻨﻬﺎ در ﺳﺎل اول ﺑﺎ روشML در ﺳﻄﺢ اﺣﺘﻤﺎل 0/01 و ﺑﺎ روش ﻫﺎی TYPEIIIوREML در ﺳﻄﺢ 0/05 ﻣﻌﻨﻲدار ﺷﺪ. ﻧﺘﺎﻳﺞ ﻧﺸﺎن داد ﻛﻪ در ﻛﻠﻴﻪ ﺳﺎل ﻫﺎ ﻣﻘﺪار ﺑﺮآوردﻫﺎ ﺑﺎ روش ML ﺑﻴﺸﺘﺮ از دو روش REML و TYPEIII ﺑﻮد. ﻫﻤﭽﻨﻴﻦ ﺑﻪ دﻟﻴﻞ ﻛﻤﺘﺮ ﺑﻮدن آﻣﺎره AIC در دو روش، REML و TYPEIII، اﻳﻦ دو روش در ﺗﺠﺰﻳﻪ داده ﻫﺎ ﻧﺴﺒﺖ ﺑﻪ روش ML ﺑﻬﺘﺮ ﻋﻤﻞ ﻛﺮدﻧﺪ.
ﻣﻜﺎن ﺗﻨﻬﺎ در ﺳﺎل دوم P<0.01( ﻣﻌﻨﻲدار ﺷﺪ. ﻫﻤﭽﻨﻴﻦ اﺛﺮ ژﻧﻮﺗﻴﭗ، ﺗﻨﻬﺎ در ﺳﺎل اول ﺑﺎ روشML در ﺳﻄﺢ اﺣﺘﻤﺎل 0/01 و ﺑﺎ روش ﻫﺎی TYPEIIIوREML در ﺳﻄﺢ 0/05 ﻣﻌﻨﻲدار ﺷﺪ. ﻧﺘﺎﻳﺞ ﻧﺸﺎن داد ﻛﻪ در ﻛﻠﻴﻪ ﺳﺎل ﻫﺎ ﻣﻘﺪار ﺑﺮآوردﻫﺎ ﺑﺎ روش ML ﺑﻴﺸﺘﺮ از دو روش REML و TYPEIII ﺑﻮد. ﻫﻤﭽﻨﻴﻦ ﺑﻪ دﻟﻴﻞ ﻛﻤﺘﺮ ﺑﻮدن آﻣﺎره AIC در دو روش، REML و TYPEIII، اﻳﻦ دو روش در ﺗﺠﺰﻳﻪ داده ﻫﺎ ﻧﺴﺒﺖ ﺑﻪ روش ML ﺑﻬﺘﺮ ﻋﻤﻞ ﻛﺮدﻧﺪ.
کلیدواژه ها
Title
Comparison of different methods to estimate variance using a mixed models in stability analysis
Authors
Abstract
Abstract
In order to Comparison of different methods to estimate variance using a mixed model, the study was performed at Gachsaran, kohdasht and Ilam agricultural research stations during the four years. Combined analysis of variance was done using the mixed model. Then, variances effects based on TYPEIII, Maximum Likelihood (ML) and Restricted Maximum Likelihood (REML) was obtained using the mixed procedure. Results indicated that the effect of location in the 4 years was not significant. The effect of genotype × location was only significant in the second year. The genotype effect was only significant in the first year for ML, the 0.01 probability level and for the TYPEIII and REML, the 0.05 probability level. Results showed that in the all years, the estimates obtained by ML method was higher than the estimates acquired by REML and TYPEIII methods. Furthermore, because the AIC statistic in REML and TYPEIII methods was lower than ML method, these two methods of data analysis are better than the ML method.
In order to Comparison of different methods to estimate variance using a mixed model, the study was performed at Gachsaran, kohdasht and Ilam agricultural research stations during the four years. Combined analysis of variance was done using the mixed model. Then, variances effects based on TYPEIII, Maximum Likelihood (ML) and Restricted Maximum Likelihood (REML) was obtained using the mixed procedure. Results indicated that the effect of location in the 4 years was not significant. The effect of genotype × location was only significant in the second year. The genotype effect was only significant in the first year for ML, the 0.01 probability level and for the TYPEIII and REML, the 0.05 probability level. Results showed that in the all years, the estimates obtained by ML method was higher than the estimates acquired by REML and TYPEIII methods. Furthermore, because the AIC statistic in REML and TYPEIII methods was lower than ML method, these two methods of data analysis are better than the ML method.
Keywords
Combined analysis of variance, Genotype × environment interaction, Grass pea, Mixed models