Analysing the genetic architecture of yield and yield-related traits in Kodo Millet (Paspalum scrobiculatum L.) through genetic parameters, correlation, and path analysis

Introduction

The small tropical millet known as Kodo millet (Paspalum scrobiculatum L.), which is indigenous to India, is grown for its seed and fodder. It is a member of the Poaceae family's Paniceae tribe and is also referred to as ditch millet, rice grass, and cow grass. The genome size of this autogamous, tetraploid (2n = 4x = 40) C4 crop species is 1.91-1.98 pg (Burton, 1940) [12] . While P. scrobiculatum var. commersonii is the wild variety native to Africa, P. scrobiculatum var. scrobiculatum is produced as a significant crop in India. Producing 450-900 kg of seed per hectare, Kodo millet is extremely hardy, drought-tolerant, and able to flourish in marginal soils where other crops could fail (Mall and Tripathi, 2016) [13] . In the Philippines, Indonesia, Vietnam, Thailand, and West Africa, among other tropical and subtropical regions, it is commonly found in moist environments. Kodo millet is high in lecithin, easily digested, and great for fortifying the nervous system (Hegde et al., 2004) [14] . One of the earliest domesticated plants, millets have been cultivated for thousands of years, providing nourishment for prehistoric societies. Because of their high nutritional content (protein, fibre, micronutrients, and phytochemicals), millets are frequently referred to as "Nutri-Cereals". Unlike other cereals, they have a low carbon and water footprint and can thrive on very poor soils under harsh, dry conditions with little input (ITBP Millet Book, 2023). Protein, fibre, and vital micronutrients including iron, calcium, and potassium are all found in Kodo millet. As a rich source of proteinaceous meals, minerals, and vitamins, Kodo millet can provide several vital ingredients needed for both human and animal body growth, maintenance, and production. Its seeds are eaten in many ways, including thin porridge and cooked rice. The Kodo millet flower's cleistogamous nature has made crop development through hybridisation challenging. Since it enables the discovery of advantageous genotypes, having an adequate collection of Kodo millet germplasm is essential for crop growth. Crop improvement depends on having a good collection of Kodo millet germplasm, and analysing it can assist in identifying desirable genotypes. To improve Kodo millet, genetic variability is essential to comprehend crop genetic behaviour in response to selection and their mode of inheritance across generations; genetic parameters such as phenotypic coefficient of variation (PCV), genotypic coefficient of variation (GCV), heritability, and genetic advance as a percentage of the mean will be computed (Anuradha et al., 2020) [1] . In Kodo millet, these traits can be utilised as selection criteria to optimise seed yield per plant. The development and control of genetic variability become crucial to crop breeding in any crop, especially in crops like Kodo millet, where the available genetic diversity is extremely limited, since genetic variability is necessary for any agricultural improvement program. Genotypes with favourable traits that could be used in the breeding program are identified as a result of this study. Understanding the relationships among yield and its component traits is equally important for developing efficient selection strategies. Correlation coefficient analysis provides information on the degree and direction of association among different characters, while path coefficient analysis partitions these associations into direct and indirect effects, thereby revealing the actual contribution of each trait towards seed yield. Such analyses enable breeders to identify key yield-contributing traits and formulate more effective selection criteria for crop improvement (Wright, 1921; Dewey and Lu, 1959) [16,3] . Considering the limited information available on the variability, diversity, morphological characteristics, and population structure of Kodo millet germplasm, the present investigation was undertaken to characterize and evaluate a diverse set of accessions (Anuradha et al., 2020) [1] . The findings of this study are expected to contribute to the identification of superior genotypes and provide a scientific basis for their utilization in future breeding programmes aimed at enhancing productivity, adaptability, and yield stability.

Materials and Methods

Planting location and experimental design

Field trials were carried out during the kharif season in 2025. The study was conducted at the SIF (Student Instructional Farm) of the Department of Genetics and Plant Breeding, Chandra Shekhar Azad University of Agriculture and Technology, Kanpur 208002, Uttar Pradesh. The experimental site is located at an altitude of 126 m above the mean sea level. Due to its humid subtropical environment, Kanpur experiences extremely low winter temperatures with the average range between 7°C-22°C and high summer temperatures ranging between 32 °C-43°C. With a few showers in winter, Kanpur receives 80% of its total rainfall during the monsoon, with average rainfall ranging around 898 mm.

Genotypes and traits evaluated

The material collected for the current experiment includes 29 Kodo millet genotypes and 2 check varieties. The checks used are GPUK3 and JK41, which were all collected from the Indian Institute of Millet Research, Rajendranagar, Hyderabad, Telangana. The list of all genotypes considered for the study is listed in Table 1. The total number of genotypes included in checks represents a broad genetic base and varies in their adaptation to different agro-climatic regions. The observations on the following morphological traits (Days to 50% flowering-DF50; Plant height (cm)-PH; Flag leaf blade length (cm)-FLL; Flag leaf blade width (cm)-FLW; Peduncle length (cm)-PDL; Panicle length (cm)-PNL; Thumb receme length (cm)-TRL; Raceme number above the panicle-RNAP; Number of basal tiller plant-NBT; Number of internode per plant-IND/P; Days to 50% maturity (DM); Green fodder weight (g)-WFT; Dry fodder weight (g)-DFT; 1000 Seed weight (g)-1000 SW; Seed yield/plant (g)-SY/P, Protein analysis (%)-Protein). Five plants of each genotype from each replication were randomly selected and tagged. Data of five plants were averaged replication-wise, and mean data were used for statistical analysis.

Statistical analysis

Using the techniques outlined by Singh and Chaudhary (1985) [7] , estimates of genotypic coefficient of variation (GCV), phenotypic coefficient of variation (PCV), broad-sense heritability (H²), and genotypic and phenotypic correlation coefficients were calculated. The methods of Johnson et al. (1955) [4] and Brim et al. (1959) [2] were used to determine genetic advance (GA) and genetic advance as a percentage of the mean (GAM). Correlation coefficient (Karl Pearson, 1912): Simple phenotypic correlation (r2) among studied traits was estimated according to Searle (1961). Using seed yield per plant as the dependent variable and its component qualities as independent variables, path coefficient analysis was conducted in accordance with the methods of Dewey and Lu (1959) [3] to distinguish direct from indirect effects. R statistical software (Version 4.5.0; R Core Team, 2025) [11] with pertinent packages like metan (Olivoto and Lúcio, 2020) [9] , agricolae (de Mendiburu, 2021), and variability (Lê et al., 2008) [8] was used for all statistical analyses, including variability parameters, correlation analysis, and path analysis.

Results

The data generated from these 31 Kodo millet genotypes were subjected to appropriate statistical analyses to derive meaningful conclusions regarding the nature and magnitude of genetic variability in the material.

Genetic Variability Parameters

In the context of trait characterisation, estimating genetic parameters such as coefficient of variation, heritability, and genetic advancement is crucial to creating high-yielding genotypes. Both the amount of the heritable part of variation and the degree of variance for a trait in a germplasm accession was examined. The coefficient of variation is a useful statistical tool for estimating the range of variability found in different demographic characteristics. However, the genetic coefficient of variation by itself cannot be used to determine the degree of heritable variation.

Genotypic coefficient of variation (GCV) and Phenotypic coefficient of variation (PCV)

The high estimates of GCV and PCV were observed for protein content (47.321 and 48.318), followed by dry fodder weight (24.710 3.

In contrast, seed yield per plant exhibited negative and nonsignificant association with raceme number above panicle (-0.1070) and dry fodder weight (-0.0210), and days to maturity (-0.0280) in Table 3. The phenotypic correlation coefficients between seed yield and its constituent variables. Days to 50 per cent flowering exhibited positive but non-significant association with panicle length (0.1370), protein content (0.1210), seed yield per plant (0.1170), 1000 seed weight (0.0420), dry fodder weight (0.0330) and number of basal tillers per plant (0.0310). However, negative and non-significant association was observed with number of internodes per plant (-0.1650), plant height (-0.1330), green fodder weight (-0.1100), peduncle length (-0.1100), thumb raceme length (-0.1060), raceme number above panicle (-0.0810), flag leaf blade width (-0.0520) and flag leaf blade length (-0.0330) in Table 3. Days to maturity exhibited positive but non-significant association with thumb raceme length (0.1580), number of basal tillers per plant (0.1690), number of internodes per plant (0.0540), green fodder weight (0.0400) and protein content (0.0810).

Negative but non-significant association was observed with seed yield per plant (-0.0280), dry fodder weight (-0.1470), 1000 seed weight (-0.0410), panicle length (-0.0080), peduncle length (-0.1060), raceme number above panicle (-0.0100), flag leaf blade length (-0.1960), flag leaf blade width (-0.0130) and plant height (-0.1260). 1000 seed weight exhibited positive but non-significant association with seed yield per plant (0.0780). Positive but nonsignificant association was observed with panicle length (0.1020), peduncle length (0.0870), thumb raceme length (0.0680), plant height (0.0670) and number of basal tillers per plant (0.0510). In contrast, significant negative association was recorded with dry fodder weight (-0.3146**). Negative but non-significant association was observed with green fodder weight (-0.0740), flag leaf blade length (-0.0490) and days to 50 per cent flowering (-0.0410) in Table 3. Protein content exhibited positive but nonsignificant association with seed yield per plant (0.1500). Significant positive association was observed with flag leaf blade width (0.2379*). Positive but non-significant association was observed with plant height (0.1340), days to 50 per cent flowering (0.1210), thumb raceme length (0.1180), peduncle length (0.0600) and number of internodes per plant (0.0440). However, negative but non-significant association was observed with green fodder weight (-0.0560), dry fodder weight (-0.0750), raceme number above panicle (-0.0850) and panicle length (-0.0320) in Table 3.

Phenotypic Path Coefficient Analysis

Path coefficient analysis was carried out at genotypic and phenotypic levels considering seed yield as the dependent trait. Path coefficient analysis was used to partition the correlation coefficients between seed yield per plant as the dependent variable and its component traits into direct and indirect effects because the genetic architecture of economic yield must be resolved with the genetic contribution of all characters influencing it directly and indirectly. Table 5 presents the direct and indirect effects of yield components on seed yield at phenotypic and genotypic levels respectively. Direct effect of yield contributing characters on seed yield per plant. In 31 genotypes of Kodo millet, path coefficient analysis was carried out to estimate the magnitude of direct and indirect effects of different traits on seed yield per plant (SY/P). Among the traits under study, flag leaf length had the highest positive direct effect (0.290) on seed yield followed by panicle length (0.262), peduncle length (0.226), flag leaf width (0.111), protein content (0.109), number of basal tillers per plant (0.080), raceme number above panicle (0.076), dry fodder weight (0.068), and number of internodes per plant (0.056). These traits contributed directly to the increase in seed yield and can therefore be considered as important selection criteria in breeding programs for yield improvement. PDL showed highest positive correlation with seed yield (0.2811) followed by FLL (0.2759), FLW (0.2275) and PNL (0.2176) in Table 4. Their positive associations with seed yield were primarily due to large positive direct effects, which indicate a true causal relationship. Thus, direct selection for these traits is expected to be effective for the improvement of seed yield. In contrast, plant height (PH) was positively correlated with seed yield (0.149) although it had a negative direct effect (-0.015), indicating that the relationship with seed yield was mainly through indirect effects via peduncle length, flag leaf traits, panicle length and protein content. The thumb raceme length also had a positive correlation (0.146) but a very low direct effect (0.009), indicating that it made a more indirect contribution to yield. DM, DF50 and green fodder yield (WFT) showed low direct effects and weak association with seed yield. RNAP had a positive direct effect but had a negative correlation with seed yield (-0.107), indicating strong adverse indirect effects through other component traits. The residual effect (0.7427) indicated that the traits considered in the present study accounted for about 44.84% (1-0.7427) of the total phenotypic variation in seed yield in Table 4. This implies that other factors not considered in the analysis also influenced expression of yield. The results showed that the most important yield contributing traits are flag leaf length, peduncle length, panicle length, flag leaf width, protein content and number of basal tillers per plant. Therefore, these characters should be given more importance during selection for the development of high-yielding genotypes of Kodo millet.

Discussion

One of our own special superfoods is millet, and the health advantages of millet are also demonstrated by the reduced rates of illness and longer lifespans of our ancestors who, a few generations ago, ate millet. Millets present a strong argument for boosting biodiversity and diversifying the food seed supply since they are adaptable to a wide range of climates and cropping patterns. Because millets can adapt to climate change and guarantee sustainable seed production with few inputs, they are the most dependable food crops for humanity in the event of climate change, particularly for resource-poor dryland farmers worldwide. Kodo millet is one of the most important crops that can survive drought because of its high nutritional value, dietary fibre, and antioxidant properties. It has a significant production potential in marginal and low-fertility soils with a variety of environmental conditions. Even under typical storage conditions, Kodo millet seeds have good storage properties and can be stored for several years without experiencing pest infestations. As a result, Kodo millet is known as the "famine seed" in some parts of the country. The success of any crop improvement programme depends largely on the extent of genetic variability present in the breeding material. Although the range provides a preliminary indication of variability, the magnitude of variation is more accurately assessed through phenotypic coefficient of variation (PCV) and genotypic coefficient of variation (GCV). In the present investigation, PCV estimates were higher than the corresponding GCV estimates for all the characters studied, indicating the influence of environmental factors on trait expression. However, the differences between PCV and GCV were relatively narrow for traits such as protein content, dry fodder weight and panicle length, suggesting that these traits were less affected by environmental fluctuations and that the observed variability was predominantly genetic in nature. Among the sixteen characters studied, the highest PCV and GCV values were recorded for protein content (48.318 and 47.321, respectively), followed by dry fodder weight (26.905 and 24.710) and panicle length (23.099 and 20.430). High estimates of both PCV and GCV indicate the presence of substantial genetic variability and provide greater scope for improvement through direct selection. The high variability observed for these traits may be attributed to the diverse genetic constitution of the genotypes included in the study. Similar findings were reported by Subramanian et al. (2010) [30] in Kodo millet, Anuradha et al. (2017) [17] in little millet, Sharma et al. (2018) [18] in pearl millet, Singamsetti et al. (2018) [23] in finger millet, and Vikram et al. (2020) [29] in barnyard millet. Moderately high PCV and GCV values were observed for flag leaf blade length (14.259 and 10.054), thumb raceme length (13.483 and 7.489), seed yield per plant (13.436 and 5.774), raceme number above panicle (16.251 and 6.830), and number of basal tillers per plant (9.446 and 5.523). These results indicate the existence of moderate variability and suggest that improvement through selection may be possible for these traits. Low PCV and GCV estimates were recorded for days to 50 per cent flowering (2.424 and 1.451), days to maturity (4.487 and 2.071), number of internodes per plant (9.215 and 3.179), green fodder weight (8.700 and 4.010), and 1000-seed weight (6.929 and 3.093). Low variability for these traits indicates a relatively narrow genetic base and limited scope for improvement through simple selection. Similar observations were reported by Ganapathy et al. (2011) [20] in finger millet and Ayesha et al. (2019) [21] in foxtail millet. The overall results indicated the existence of considerable variability among the Kodo millet genotypes, particularly for protein content, dry fodder weight and panicle length, which could be effectively exploited in future breeding programmes. The coefficient of variation alone does not provide information regarding the transmissibility of a trait from one generation to the next. Therefore, heritability estimates are essential for determining the proportion of phenotypic variation attributable to genetic causes. Heritability, when considered along with genetic advance, provides a reliable indication of the effectiveness of selection. In the present study, high heritability estimates were observed for protein content (95.914%), dry fodder weight (84.349%), and panicle length (78.226%). High heritability indicates that these traits are predominantly governed by genetic factors and are less influenced by environmental conditions. Therefore, phenotypic selection for these traits would be highly effective. Similar findings were reported by John (2006) [22] in finger millet, Sharma et al. (2018) [18] in pearl millet, and Singamsetti et al. (2018) [23] in finger millet. Moderate heritability was observed for flag leaf blade length (49.712%), days to 50 per cent flowering (35.859%), number of basal tillers per plant (34.191%), thumb raceme length (30.847%), and plant height (26.234%). Moderate heritability suggests that both genetic and environmental factors contribute to the expression of these traits. Consequently, selection based solely on phenotype may not always be reliable, and evaluation across environments may be necessary. Low heritability estimates were recorded for days to maturity (0.213%), peduncle length (10.863%), number of internodes per plant (11.901%), raceme number above panicle (17.661%), seed yield per plant (18.469%), green fodder weight (21.241%), flag leaf blade width (21.981%), and 1000-seed weight (19.929%). Low heritability indicates a greater influence of environmental factors on the expression of these traits and suggests that direct selection may be less effective. The results revealed that protein content, dry fodder weight and panicle length possessed high heritability and therefore offer considerable potential for genetic improvement through selection. Heritability estimates alone do not provide complete information regarding the expected progress from selection. Johnson et al. (1955) [4] emphasized that heritability coupled with genetic advance provides a more reliable estimate of the effectiveness of selection. Genetic advance as percent of mean (GAM) indicates the expected improvement in a trait under selection and helps in understanding the nature of gene action controlling the trait. In the present investigation, high genetic advance as percent of mean was observed for protein content (95.469%), dry fodder weight (46.751%), and panicle length (37.223%). These traits also exhibited high heritability, indicating the predominance of additive gene action. Therefore, substantial improvement in these traits can be achieved through simple phenotypic selection. Similar findings were reported by Talaware et al. (2017) [24] in pearl millet, Singamsetti et al. (2018) [23] in finger millet, and Vikram et al. (2020) [19] in barnyard millet. Moderate genetic advance as percent of mean was observed for flag leaf blade length (14.602%), thumb raceme length (8.568%), number of basal tillers per plant (6.653%), raceme number above panicle (5.913%), and seed yield per plant (5.112%). Moderate GAM coupled with moderate heritability suggests the involvement of both additive and non-additive gene effects in the inheritance of these traits. Improvement through selection is possible, although progress may be relatively slower. Low genetic advance as percent of mean was recorded for days to 50 per cent flowering (1.790%), days to maturity (1.968%), plant height (3.979%), flag leaf blade width (4.546%), peduncle length (3.581%), number of internodes per plant (2.259%), green fodder weight (3.807%), and 1000-seed weight (2.845%). Low GAM associated with low to moderate heritability indicates the predominance of non-additive gene action and environmental influence. Therefore, direct selection for these traits may not be highly effective, and alternative breeding approaches such as hybridization followed by selection in later generations may be more useful. The combination of high heritability and high genetic advance observed for protein content, dry fodder weight and panicle length suggests that these traits are primarily governed by additive gene effects and can be effectively improved through selection. These traits may therefore serve as important selection criteria in future Kodo millet breeding programmes aimed at enhancing productivity and nutritional quality. At the phenotypic level, seed yield per plant exhibited significant positive association with flag leaf blade length (0.2759**), flag leaf blade width (0.2275*), peduncle length (0.2811**), and panicle length (0.2176*). The magnitudes of these phenotypic correlations were generally lower than the corresponding genotypic correlations, which is a commonly observed phenomenon attributable to environmental masking effects on the expression of the traits. Among the significant traits, peduncle length and flag leaf blade length showed comparatively higher positive association with seed yield at the phenotypic level, suggesting their importance in translocation of photosynthates and development of the reproductive sink. Similar results were reported by Priyadharshini et al. (2011) [25] in finger millet. Plant height (0.1489), thumb raceme length (0.1456), protein content (0.1499), number of basal tillers per plant (0.1103), green fodder weight (0.0599), and 1000 seed weight (0.0781) showed positive but non-significant phenotypic association with seed yield per plant, indicating relatively lower direct influence on yield under the present environmental conditions. In contrast, seed yield per plant showed negative non-significant association with raceme number above panicle (-0.1067) and dry fodder weight (-0.0213) at the phenotypic level, consistent with the genotypic level findings. Similar positive associations of grain yield with flag leaf traits and panicle-related characters were documented by Suryanarayana and Sekhar (2018) [26] in little millet and Kodo millet. The correlation coefficient analysis at the phenotypic level, twelve out of fourteen characters had positive direct effects on seed yield per plant, while days to maturity (-0.031) and green fodder weight (-0.014) exerted negative direct effects. The highest positive direct effect at the phenotypic level was recorded for flag leaf blade length (0.290), followed by panicle length (0.262), peduncle length (0.226), flag leaf blade width (0.111), protein content (0.109), days to 50% flowering (0.114), number of basal tillers per plant (0.080), raceme number above panicle (0.076), dry fodder weight (0.068), number of internodes per plant (0.056), 1000 seed weight (0.048), and thumb raceme length (0.009). The high positive direct effects of flag leaf blade length, panicle length, and peduncle length at the phenotypic level were consistent with their significant positive correlations with seed yield and their biological roles in assimilate production and transport. The positive direct contribution of flag leaf blade length confirmed that efficient photosynthetic area plays a critical role in grain development in Kodo millet. These results are in agreement with the findings of Pallavi et al. (2020) [27] in foxtail millet and Parsha Hema et al. (2022) [28] in finger millet. The residual effect of 0.7427 indicated that the traits included in the phenotypic analysis accounted for a substantial proportion of the variation in seed yield, while the remaining variation was governed by other environmental and genetic factors not considered in the present analysis. The path coefficient analysis at the phenotypic level, days to 50% flowering exerted positive indirect effects on seed yield through panicle length (0.036) and protein content (0.013), while negative indirect effects were observed through peduncle length (-0.025) and number of internodes per plant (-0.009). Plant height showed positive indirect effects through peduncle length (0.082), panicle length (0.022), and protein content (0.015), suggesting that taller plants indirectly contribute to yield through better panicle development and biomass accumulation. Flag leaf blade length showed positive indirect effects through peduncle length (0.032) and number of basal tillers per plant (0.007), while negative indirect effects through panicle length (-0.040) and raceme number above panicle (-0.014) partially offset these contributions. Flag leaf blade width exerted positive indirect effects through peduncle length (0.066), protein content (0.026), and flag leaf blade length (0.027). Peduncle length showed positive indirect effects through flag leaf blade length (0.042) and flag leaf blade width (0.030), while negative indirect effects through raceme number above panicle (-0.016) reduced its overall contribution. Panicle length contributed positive indirect effects through flag leaf blade length (0.044) and green fodder weight (0.030), while number of basal tillers per plant showed positive indirect effects mainly through panicle length (0.026) and flag leaf blade length (0.025). Overall, the phenotypic path coefficient analysis confirmed that flag leaf blade length, flag leaf blade width, peduncle length, panicle length, and protein content were the most important characters for selection in Kodo millet yield improvement, a finding consistent with observations of Kumar et al. in Kodo millet.

Conclusion

Based on the current study's findings, it can be said that there is significant genetic variation among the assessed Kodo millet genotypes for yield and related characteristics, offering plenty of room for genetic improvement via selection. Protein content and dry fodder weight showed strong genetic advance as a percentage of the mean and high heritability, demonstrating that direct phenotypic selection can successfully improve these qualities and indicating the predominance of additive gene action. Seed yield per plant, on the other hand, demonstrated poor genetic progress (5.11%) and low heritability (18.47%), suggesting a stronger impact of environmental factors and restricting the efficacy of direct selection for yield increase. Consequently, it seems that indirect selection using component qualities is a more successful breeding approach. Flag leaf blade length, flag leaf blade width, panicle length, peduncle length, and thumb raceme length were all positively correlated with seed yield, according to correlation analysis, underscoring their significance as selection criteria in breeding programs. Additionally, route coefficient analysis showed that raceme number above panicle and thumb raceme length had significant direct influences on seed yield, highlighting the importance of raceme architecture in yield determination. Therefore, while concurrently increasing protein content and biomass output, breeding efforts should concentrate on improving these yield-contributing qualities. The production of highyielding, nutritionally superior, and agronomically desirable kodo millet genotypes appropriate for sustainable farming and upcoming crop improvement initiatives would be made easier by such an integrated selection technique.

References

1. Anuradha N, Patro TSSK, Triveni U, Rao PJ, Rani YS, Priya PK. Assessment of genetic variability estimates in kodo millet (Paspalum scrobiculatum L.). International Journal of Chemical Studies. 2020;8(5):1907-1909.
2. Brim CA, Johnson HW, Cockerham CC. Multiple selection criteria in soybeans. Agronomy Journal. 1959;51:42-46.
3. Dewey DR, Lu KH. A correlation and path coefficient analysis of components of crested wheatgrass seed production. Agronomy Journal. 1959;51:515-518.
4. Johnson HW, Robinson HF, Comstock RE. Estimates of genetic and environmental variability in soybeans. Agronomy Journal. 1955;47:314-318.
5. Mahalanobis PC. On the generalized distance in statistics. Proceedings of the National Institute of Sciences of India. 1936;2:49-55.
6. Rao CR. Advanced statistical methods in biometrical research. New York: John Wiley & Sons; 1952.
7. Singh RK, Chaudhary BD. Biometrical methods in quantitative genetic analysis. New Delhi: Kalyani Publishers; 1985.
8. Lê S, Josse J, Husson F. FactoMineR: An R package for multivariate analysis. Journal of Statistical Software. 2008;25(1):1-18.
9. Olivoto T, Lúcio AD. metan: An R package for multienvironment trial analysis. Methods in Ecology and Evolution. 2020;11:783-789.
10. Kassambara A, Mundt F. factoextra: Extract and visualize the results of multivariate data analyses. R package version 1.0.7. 2020.
11. R Core Team. R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2025.
12. Burton GW. A cytological study of some species in the genus Paspalum. Journal of Agricultural Research. 1940;60:193198.
13. Mall TP, Tripathi SC. Millets: The nutrimental potent ethnomedical grasses. World Journal of Pharmaceutical Research. 2016;5(2):513.
14. Hegde P, Rajasekaran N, Chandra T. Effect of antioxidant properties of millet species on oxidative stress and glycemic status in alloxan-induced rats. Indian Journal of Nutritional Research. 2004;25(12):1109-1120.
15. Wright S. Correlation and causation. Journal of Agricultural Research. 1921;20:557-585.
16. Anuradha N, Patro TSSK, Divya M, Rani YS, Triveni U. Genetic variability, heritability and correlation of quantitative traits in little millet genotypes. Journal of Pharmacognosy and Phytochemistry. 2017;6(6):489-492.
17. Sharma B, Chugh LK, Sheoran RK, Singh VK, Sood M. Study on genetic variability, heritability and correlation in pearl millet germplasm. Journal of Pharmacognosy and Phytochemistry. 2018;7(6):1983-1987.
18. Vikram S, Sudhagar R, Masilamani P, Vanniarajan C. Genetic variability and association analysis in barnyard millet mutants. Electronic Journal of Plant Breeding. 2020;11(1):210-216.
19. Ganapathy S, Nirmalakumari A, Muthiah AR. Genetic variability and interrelationship analyses for economic traits in finger millet germplasm. World Journal of Agricultural Sciences. 2011;7(2):185-188.
20. International Journal of Current Microbiology and Applied Sciences. 2019;8(4):2173-2179.
21. John K. Variability and correlation studies in quantitative traits of finger millet (Eleusine coracana L. Gaertn). Agricultural Science Digest. 2006;26(1):166-169.
22. Singamsetti A, Patro TSSK, Anuradha N, Divya M. Studies on genetic variability for yield and yield attributing traits in finger millet (Eleusine coracana L. Gaertn). International Journal of Current Microbiology and Applied Sciences. 2018;7:90-95.
23. Talawar AM, Girish G, Channabasavanna AS, Kitturmath MS. Studies on genetic variability, correlation and path analysis in pearl millet (Pennisetum glaucum L.) germplasm lines. 2017.
24. Priyadharshini C, Nirmalakumari A, Joel AJ, Raveendran M. Genetic variability and trait relationships in finger millet (Eleusine coracana L. Gaertn.) hybrids. Madras Agricultural Journal. 2011;98(1-3):18-21.
25. Suryanarayana L, Sekhar D. Studies on genetic variability, character association and path analysis in little millet (Panicum sumatrense L.) genotypes. The Pharma Innovation Journal. 2018;7(7):908-910.
26. Pallavi NL, Venkatesh R, Ram BJ, BG S. Studies on correlation and path coefficient analysis in foxtail millet [Setaria italica (L.) Beauv.]. International Journal of Chemical Studies. 2020;8(6):1941-1946.
27. Parsha Hema, Elangovan SP, Yadav BN. Study of genetic variability and character association in elite lines of finger millet [Eleusine coracana (L.)] suitable for Eastern Plain Zone of Uttar Pradesh. International Journal of Environment and Climate Change. 2022;12(12):438-447.
28. Thakur J, Kanwar RR, Kumar P, Salam JL, Kar S. Studies of genetic parameters for yield and yield attributing traits of kodo millet (Paspalum scrobiculatum L.). International Journal of Current Microbiology and Applied Sciences. 2018;7(9):278-287.
29. Subramanian A, Nirmalakumari A, Veerabadhiran P. Trait based selection of superior kodo millet (Paspalum scrobiculatum L.) genotype. Electronic Journal of Plant Breeding. 2010;1(4):852-855.