STEPWISE REGRESSION AND PRINCIPAL COMPONENT ANALYSES FOR QUANTITATIVE TRAITS OF RAPESEED GENOTYPES AT DIFFERENT SOWING DATES

The present research was done to assess the best selection criteria for yield improvement in rapeseed (Brassica napus L.) using stepwise regression and principal component analyses at different sowing dates. All the traits except 1000-seed weight were significantly affected by sowing dates. The results of stepwise regression analysis revealed that seeds per pod had an important role at the first and second sowing dates, but at the third and fourth sowing dates, pods per plant and days to flowering were more important than other yield components for a seed yield prediction model. On the basis of a cumulative percent of variation, three principal components (PCs) were determined for each sowing date. The cumulative percentages of variation for three PCs at the first to fourth sowing dates were 0.97, 0.96, 0.89 and 0.95, respectively. At the first sowing date, the first principal component (PC1) had high positive and negative PC loading values for the studied traits such as days to flowering, days to the end of flowering, duration of flowering, pods per plant and harvest index. Therefore, there was high variation in these traits among the genotypes. PC2 of the first sowing date had also high PC loadings for pods on the main raceme, seeds per pod, 1000-seed weight, biological and seed yields, therefore the correlation of these traits with this PC will be high. In PC3 of the first sowing date, height, pods on the main raceme and pods per plant had the high value of PC loadings. Based on stepwise regression analysis, seeds per pod at the first and second sowing dates and days to flowering and pods per plant at the third and fourth sowing dates had an important role for improving seed yield.


Introduction
Rapeseed (Brassica napus L.) is known as the second most important source of high quality vegetable oil in the world.Its high quality is related to the lowest amount of saturated fatty acids in comparison to other oilseed crops and moderate substance of poly-unsaturated fatty acids (Stamer et al., 1999).Seed yield, as the most important quantitative characteristic, will be a result of genotype, environment and genotype-environment interaction effects (Sharief and Keshta, 2002;Marjanovic-Jeromela et al., 2009).Due to remarkable effects of sowing dates on component characters and seed yield, the most important traits such as phenological and morphological traits, plant height and pods were significantly decreased at late planting dates and therefore seed yield was decreased (Burhan, 2007;Siadat and Hemayati, 2009).Correlation analysis helps researchers to discriminate a significant relationship between traits.Non-important traits in the regression model will be omitted via a stepwise regression analysis, therefore the most important traits which have considerable effects on the dependent variable will be verified.Most of the traits which were selected through the stepwise regression model can be used as selection criteria for indirect selection in a breeding program (Aytac and Kinaci, 2008;Ogrodowczyk and Warzyniak, 2004;Sabaghnia et al., 2010;Wiliams et al., 1990).Chango and McVetty (2001) reported that total dry matter and harvest index had a significant correlation with grain yield in spring type rapeseed genotypes.Khan et al. (2006) concluded that the number of pods per plant was significantly correlated with grain yield in rapeseed.A significant correlation was determined between pod number per plant and seed yield in species of B. napus and B. juncea (Akbar et al., 2007;Marjanovic-Jeromela et al., 2009).This demonstrated that among seed yield component characters, the number of pods had the greatest effect, whereas seed per pod and seed weight had a small influence on seed yield (Farshadfar and Farshadfar, 2008;Khalily et al., 2010;Belete, 2011).Different statistical techniques have been applied in modeling crop yield, including regression, correlation, path analysis, principal components, factor and cluster analyses (Khaliq et al., 2001;Sagir et al., 2004;Mahasi et al., 2006).The principal component analysis (PCA) is concerned with explaining the variance-covariance structure through a few linear combinations of the original variables.The most important objectives of PCA are data reduction and interpretation.Although all estimated principal components (PCs) are required to reproduce the total system variability, often much of this variability can be accounted for a small number of uncorrelated PCs (Leilah and Al-Khateeb, 2005;Saporta and Niang, 2009).Mohammadjani Asrami et al. (2014) applied the PCA to determine the diversity and grouping of rapeseed genotypes based on the quantitative characteristics.The PCA revealed that the first three traits corresponded to the whole percentage of the variance in the dataset.The first three main PCs were detected from the complicated components, and the total cumulative variance of these three factors was 87.5.The first PC was related to plant height, protein and oil percents.The second PC included high vector loadings for pod length, number of seeds per pod and 1000-seed weight.The third PC distinguished variables that were related only to the percent of glucosinolate.
The objectives of the present study were to detect the sowing date effects on yield correlated traits and also to estimate stepwise regression and principal component analyses for different sowing dates to recognize suitable selection criteria for the breeding program of spring rapeseed genotypes.

Material and Methods
In order to assess the sowing date effects on yield associated component characters and results of stepwise regression and principal component analyses in rapeseed genotypes, a split-plot experiment in a randomized complete block design with three replications was conducted at Biekol Agriculture Research Station, located in Neka, Iran (53˚, 13 ′ E longitude and 36 ˚ 43 ′ N latitude, 15 m above sea level) during the 2011-2012 cropping season.Four sowing dates including the 18 th and the 31 st of October as well as the 5 th and the 15 th of November were considered as main plots and the cultivars including L10 (19HSAR), Sarigol, L7(SLMR308), L4(SLM308) and Hyola401 were considered as sub-plots.The soil was grouped as a deep loam soil (Typic Xerofluents, USDA classification) composed of an average of 280 g clay kg -1 , 560 g silt kg -1 , 160 g sand kg -1 , and 22.4 g organic matter kg -1 with a pH of 7.3.Each sub-plot was consisted of four rows 5 m long and 30 cm apart.The distance between plants on each row was 5 cm resulting in approximately 400 plants per plot, which was adequate for statistical analyses.Crop management factors including land preparation, fertilizer application, and weeds were controlled based on local area recommendation.All the plant protection measures were applied to make the crop free from hazarding agents.The data were recorded based on ten randomly selected plants of each treatment of each replication for days to flowering, days to maturity, plant height, pods per plant, seeds per pod and 1000-seed weight.Twenty plant samples selected from the middle part of each plot were used to measure the traits including biological yield and harvest index.Seed yield of two middle rows of each plot, excluding the border area, was measured at physiological maturity and yield was adjusted to 12.5% seed moisture content.Analysis of variance of the split-plot experiment was done for all the traits (Steel et al., 1997).All the analyses were performed using MS-Excel and SAS software version 9 (SAS INSTITUTE INC, 2004).

Analysis of variance
The traits including days to flowering, duration of flowering and days to maturity, plant height, yield components except 1000-seed weight, seed and biological yields were significantly affected by sowing dates (Table 1).The genotypes also had a significant genetic difference for all the traits.Significant interaction effects of sowing dates and genotypes were detected for phonological traits, pods per plant and seeds per pod.Significant sowing date effects were reported for phenological and morphological traits, yield components and seed yield in rapeseed (Sharief and Keshta, 2002;Siadat and Hemayati, 2009).

Stepwise regression analysis
The result of stepwise regression analysis for different sowing dates is presented in Table 2.At the first sowing date, the number of seeds per pod, as a prime important trait, was entered in the model at the first stage, and the number of pods on main raceme was the second important trait and was inserted in the model at the second stage.In general, the model for detecting of seed yield was determined as Y=-873.55+13.69X6+130.30X8, in which Y, X6 and X8 were seed yield, pods per plant and pods on main raceme, respectively.At the second sowing date, different traits had an important role for predicting seed yield, therefore the selection breeding program for improving of seed yield should be optimized based on each sowing date for the certain environment.The traits including 1000-seed weight and seeds per pod were important for the second sowing date and also were entered into the model at the first and second stages, respectively.The model for seed yield was determined as Y=1275.58+83.56X6-707.01X9, in which Y, X6 and X9 were used as seed yield, seeds per pod and 1000-seed weight, respectively.At the third sowing date, pods per plant had an important role for improving seed yield and the final model was determined as Y=-4231.33+48.33X7,wherein Y and X7 were seed yield and pods per plant, respectively.At the fourth sowing date, days to end of flowering was important for determining seed yield, and the model gained: Y=18584-104.7X2, in which Y and X2 indicated seed yield and days to end of flowering, respectively.In earlier studies (Khayat et al., 2014;Mohammadjani Asrami et al., 2014) based on stepwise regression analysis, it was reported that pods per plant and plant height had an important role for seed yield prediction in rapeseed genotypes.In stepwise regression analysis for quantitative traits of rapeseed genotypes, the number of grains per pod verified 67%, pods per plant 64% of coefficient, 1000-seed weight 72%, oil percentage 78% and the number of nodes in stem verified 80% of coefficient changes in the regression model (Baradaran et al., 2006).In the regression model conducted by Rahnamaee et al. (2007) in order to determine effective traits on bean yield, the traits including 100-seed weight, the number of grains per pod and total number of pods were entered into the model.

Principle component analysis
Principal component analysis (PCA) involves a mathematical procedure that transforms a number of correlated variables into a smaller number of uncorrelated variables called principal components (Saporta and Niang, 2009).The first principal component accounts for as much of the variability in the data as possible, and each succeeding component accounts for as much of the remaining variability as possible.PCA was performed for each sowing date using the standard procedure.On the basis of the cumulative percent of variation, three principal components (PCs) were determined at each sowing date (Table 3).The cumulative percent of variation for three PCs at the first to the fourth sowing dates were 0.97, 0.96, 0.89 and 0.95, respectively.At the first sowing date, the first principal component (PC1) had high positive and negative PC loading values for the traits including days to flowering, days to end of flowering, duration of flowering, pods per plant and harvest index, therefore the genotypes showed high variation for these traits.PC2 of the first sowing date had also high PC loadings for the traits such as pods on main raceme, seeds per pod, 1000-seed weight, biological and seed yields, therefore correlation of these traits with this PC will be high.In PC3 of the first sowing date, height, pods on main raceme and pods per plant had the high value of PC loadings.At the second sowing date, PC1 had high PC' loadings for days to flowering, days to end of flowering, days to maturity, harvest index and seed yield.Duration of flowering, plant height, pods on main raceme, seeds per pod, 1000-seed weight and biological yield had high PC loadings in PC2 of the second sowing date.In PC3 of the third sowing date, days to end of flowering, duration of flowering, plant height, seeds per pod, 1000-seed weight and biological yield had high positive and negative PC loadings.At the third sowing date, days to flowering, days to end of flowering, days to maturity, plant height, harvest index and seed yield had the high value of PC1 loadings.Also, in PC2 of the third sowing date, pods per plant, seeds per pod, biological and seed yields had high PC loadings.The traits including days to end of flowering, duration of flowering had high PC loadings in PC3 of the third sowing date.At the fourth sowing date, days to flowering, days to end of flowering, duration of flowering, days to maturity, pods per plant, seeds per pod, harvest index and seed yield had the high value of PC1 loadings (Table 3).In PC2 of the fourth sowing date, plant height, pods on main raceme, pods per plant, 1000-seed weight and biological yield had the high value of PC loadings.In PC3 of the fourth sowing date, plant height, biological and seed yields proved to be the most important variables since they had high positive and negative PC loading values.Mohammadjani Asrami et al. ( 2014) applied PCA to assess the diversity and grouping of rapeseed genotypes based on the characteristics and parameters.The PCA shows that the first three traits correspond to the whole percentage of the variance in the dataset.The first three main PCAs were extracted from the complicated components, the total cumulative variance of these three factors amounted to 87.5%.In their study, the PCA grouped the estimated canola variables into three main components, of which PCA1 accounted for about 37.236% of the variation; PCA2 for 33.26%;PCA3 for 17.038%.The first PCA was related to plant height, percent of protein and percent of oil, whereas the second PCA was related to pod length, number of seed in pod and 1000-seed weight.The third PCA contrasted variables that were related solely to the percent of glucosinolate.

Table 1 .
Analysis of variance for phonological traits, yield components and seed yield.

Table 2 .
The results of stepwise regression analysis of 12 studied traits.

Table 3 .
Principal component analysis for 12 studied traits in rapeseed genotypes under four sowing dates.