Results
Association among various studied characters
The basic statistics of different studied traits revealed considerable amount of variability among 159 exotic cotton genotypes (Table 1). Simple correlation coe- fficient revealed some important associations among the studied traits table 2.
 

Days to first square exhibited positive significant association with days to first flower and nodes to first fruiting branch while it had significant negative association with fiber fineness and negatively and non significantly correlated with plant height, monopodia per plant, sympodia per plant, boll weight and staple length. Days to first square showed positive significant association towards nodes to 1st fruiting branch and had negative non-significant correlation with plant height, monopodia, sympodia per plant , boll weight and fiber fineness. The CLCuD exhibited highly significant negative association with plant height while showed negative non-significant correlation with monopodia per plant and staple length. The plant height had highly significant correlation with monopodia per plant and sympodia per plant while showed highly significant negative association with GOT. On the other hand nodes to first fruiting branch exhibited significant negative association with fiber fineness and for staple length significant negative correlation was found with fiber fineness.

Principle Component analysis
In present set of experiment, out of total 11 (PCs) five principle component (PCS) were obtained having >1 Eigen value. All these five PCs contributed 65% in total variability amongst the upland cotton genotypes which were assessed for genetic diversity for fiber quality traits, earliness and CLCuD (Table 3). The remaining 35% of the total variability towards total variation was contributed by other components.
 

Cluster Analysis
By using cluster analysis 159 exotic genotypes were grouped into 6 clusters on the basis of various traits (Table 4). That cluster I, II, III, IV, V and VI Comprised of 25, 18, 23, 46,27 and 20 genotypes respectively (Name of genotypes not shown). The genotypes in the first cluster represent reasonable values for days to first squaring, flowering, sympodial branches per plant, staple length and fiber fineness while for other traits selection cannot be made in this cluster. The members of 2nd cluster exhibited lower values for CLCuD, fiber fineness and GOT% whilst plant height, staple length and sympodia per plant showed higher values. The cluster III was characterized by genotypes having maximum GOT, staple length and fiber fineness. Cluster IV showed reasonable values for CLCuD, GOT and staple length. Cluster V was found suitable for only staple length while genotypes in cluster VI had maximum values for plant height and sympodia per plant. The first two principle components contributed almost 34.1 % in total variance. Among all clusters not a single cluster exhibited obvious separation.
    

Discussion
The information of association among various traits facilitates to initiate any breeding programs as it aids to select genotypes having superior characters (Ali et al., 2009). In this set of experiment the correlation analysis results showed significant association among various traits. The days to 1st square had significant positive correlation with days to first flowering and had positive association with nodes to 1st fruiting branch whilst it showed negative correlation with fiber fineness. CLCuD exhibited highly significant negative correlation with plant height. Plant height showed highly positive significant association with sympodia per plant and monopodia per plant while it showed significant negative association with GOT. The positive association among various studied traits is important for selection of high yielding genotypes. Farooq et al. (2013) found significant positive association among yield and yield contributing traits and also found negative correlation among CLCuD and seed cotton yield. The exploitation and maintenance of genetic resources could be made more efficient by partitioning of total variance into its components. Meanwhile it also aids in utilization of suitable germplasm for improvement of particular plant character (Pecetti et al., 1996). Principle component analysis is a powerful technique that facilitates to get appropriate parental lines thus to initiate successful breeding program (Akter et al., 2009). In present set of experiment, PC analysis grouped total variation into 5 PCs. Saeed et al. 2014 reported significant higher contribution of 1st and 2ndprinciple component towards total variability. In present experiment 1st PC was elaborated by mainly due to days to 1st square, days to first flower and CLCuD. The 2nd PC was mainly explained by the genotypes having diversity for plant height and monopodia per plant. PC III contribution towards total variance was due to diversity among genotypes for nodes to 1st fruiting branch, staple length and CLCuD. PC IV elucidated by diversity among genotypes for staple length, monopodia per plant and boll weight with positive factor loadings.

Principle component analysis eventually validated ample amount of variation for the studied traits which could be exploited for planning a successful breeding program initiative aimed at refining earliness, CLCuD and fiber quality traits. Malik et al. 2011 and Ashokkumar and Ravikesavan 2011 reported that abundant amount of diversity in colored cotton genotypes allow opportunities to characterize color cotton genotypes.

Cluster analysis also assisted to elucidate adequate amount of diversity in this set of genotypes at different growth phases. The genotypes in cluster I, II and III comprised of genotypes with better earliness and fiber quality traits. Cluster IV was characterized by genotypes having less attack of CLCuD, higher GOT %and somehow good quality of fiber traits. Cluster VI was represented by genotypes having good plant height, sympodial branches per plant and fiber fineness. Genotypes of cluster I, II and III could be exploited in new breeding programs because of their better earliness and fiber quality traits. Cluster IV  genotypes are good to utilize  for better CLCuD tolerance and GOT% whilst Cluster VI genotypes are suitable for plant height, sympodial branches per plant and fiber fineness.

The existence of inclusive diversity between clusters is of great genetic values that allow selecting the genotypes with wider genetic base for better tolerance against CLCuD and earliness traits. Ayana and Bekele et al. reported similar kind of results for grouping of germplasm into various clusters.

Conclusion

In this present set of experiment PC analysis, correlation coefficient and cluster Analysis was employed that provide facilitation in grouping of genotypes and identification of genotypes having better tolerance against CLCuD, better fiber quality and earliness trait. The information gathered through correlation, PC and cluster analysis will be fruitful to design breeding programs to obtain genotypes possessing higher fiber quality, CLCuD tolerance and earliness traits.

Materials and Methods

Plant Material & Site Characteristics             
A total of 159 genotypes were evaluated for this study carried out during the cropping seasons 2012-13 on 19th of June. The experiment was carried out at Cotton Research Institute, Faisalabad, Punjab, Pakistan.

Experimental Design, Plot Size & Cultural Practices

For each entry, plot size measured 4.572 m × 1.524 m, comprising 2 rows set 75 cm apart. Distance between plants within rows was 30 cm. Normal agronomic and cultural practices (irrigation, weeding, hoeing, and fertilizer applications) were adopted as and when required.

Measurement of the studied traits
For measuring the traits 10 representative, undamaged plants were selected in each line and marked for identification. Data were collected for nodes to 1st fruiting branch counted from zero node (cotyledonary node) to the node at which first flower had appeared, number of days to  first square and flower, plant height, monopodia and sympodia per plant, boll weight, and ginning out turn (GOT). For GOT cleaned and dry samples of seed cotton were weighed and then ginned separately with single roller electric ginning machine. The lint obtained from each sample was weighed and ginning out turn % was calculated by the following formula: Ginning outturn (%) = Weight of lint/Weight of seed cotton × 100, Fiber charac- teristics such as staple length, fiber fineness of each guarded plant were measured by using spin lab HVI-900.

Cotton leaf curl virus disease incidence (%) methodology
Cotton leaf curl virus disease incidence (%) and reaction of the cultivars was determined by using disease scale (Table 5) modified from the scale described by Akhtar et al (2010). Then percentage of CLCuD incidence was calculated by using the following formula. CLCuD incidence (%) = Sum of all disease ratings/ total number of plants ×25.
  

Table 5 Rating scale for cotton leaf curl virus disease (CLCuD) symptoms

Statistical Analysis
The average data of both the years were subjected to basic statistics, correlation analysis, cluster analysis and principal component analysis (PCA) using statistical software packages of SPSS version 19 and STATISTICA version 5.0 (Sneath and Sokal, 1973). Cluster analysis was performed using K-means clustering while tree diagram based on elucidation distances was developed by Ward’s method. First two principal components were plotted against each other to find out the patterns of variability among genotypes and association between different clusters using SPSS version 19.

Acknowledgements
This material is based upon work supported by the USDA Research Service under agreement No. 58-6402-178F. Any opinion, findings, conclusions or recommendations expressed in this publication are those of author(s) and do not necessarily reflect the views of USDA