Validating Technologies and Evaluating the Technological Level in Avocado Production Systems: A Value Chain Approach

: In agriculture, technologies support the productivity and competitiveness of production systems in value chains. In the last decade, the Colombian avocado sub-sector has expanded. However, little is known about its technological level (TL). The objectives of this study were (i) to understand the perception of value chain experts in terms of technological supplies and demands, (ii) to evaluate the TL in avocado production systems and (iii) to discover which socioeconomic characteristics impact the TL. The three stages were consultation with experts and parameterization of the TL, evaluation of the TL using multivariate methods and analysis of variables of the production system. The technological supply “By-products from seed” was of great importance, while “Branch girdling” was the least relevant. A total of 161 demands were identiﬁed, “Environmental management and sustainability” showing greater relevance. The analysis showed a low correlation between the qualiﬁcation of experts and the use of technology by producers. The postharvest supplies had the lowest frequency of use. Producers were characterized according to TL: high (34.4%), medium (47.2%) and low (18.4%). A relationship was found between the TL and some variables of the production system. The gap found should be the basis for designing science and technology policies for the avocado sub-sector in Colombia.


Introduction
Technology is a large part of daily activities, from educational processes to agricultural work [1,2].The concept of technology has been commonly applied to artifacts, tools or machinery, which are called hard technologies [3].However, knowledge and its management are also classified as technologies, which seek to facilitate, optimize or improve processes.This has been called soft technology [4].
In agriculture, technologies have been used to intensify food production [5,6], where soft technologies are implicitly linked to management practices.Some relevant technologies include selecting varieties, sustainable soil management and smart farming for resource management.Recently, information and communication technologies have entered production systems [7].In terms of technology, there are gaps and resources that give rise to sustainable and comprehensive processes [8].This information facilitates the execution of improvement actions.Although technology is always present, technology users can decide to use it or not, which is called adoption [9].
Technology can support agriculture through business models, improved production, cost management, business diversification, energy efficiency, process optimization, etc. [7,[10][11][12].However, the target group must be known to allow technology transfer as well as pertinent and efficient research [13].Some production systems, nevertheless, have technological lags as a result of low or null technological adoption, which generates

Materials and Methods
Avocado production in Colombia is distributed throughout the country.The main production areas are Tolima, Caldas, Valle del Cauca and Antioquia.Currently, there are some 13,000 avocado producers growing a wide variety of avocado, but that variety that shows the greatest growth in planted area is Hass, due to its export potential [32].
This research establishes a methodological proposal for the avocado value chain and constitutes a first approach is appraising the technological level of the industry.To respond to the objectives, the research was undertaken in three parts: (i) Consultation with experts regarding existing supplies and demands, (ii) Evaluation of the technological level and (iii) Correlation of the technological level with socioeconomic characteristics.

Consultation with Experts and Existing Demands in Avocado Production Systems
The opinion of experts in the avocado value chain (producers, researchers, technical assistants and marketers) was elucidated with an instrument built with four groups and a number of technological supplies: (i) Propagation of plant material and nursery (20), (ii) Production ( 47), (iii) Postharvest ( 25) and (iv) Management and marketing (22).Their opinion was solicited for the importance of technological supplies and demands in this value chain.The Likert scale [33] was used, with five response options: 5-very important, 4-important, 3-moderately important, 2-low importance and 1-unimportant.The participants were selected from a bibliometric analysis based on the following criteria: (i) research with data capture and (ii) validation of the research through data analysis processes and under field conditions.
Researchers with postgraduate training who reported more than 10 publications on avocado on the ScienTI platform of the Ministry of Science, Technology and Innovation, were sought.To contact marketers, a web search of the main avocado marketing agents in Colombia was carried out.Producers identified themselves through associations and social networks.Technical assistants were contacted through Linkata and social networks.Experts' responses were not categorized according to role, due to the relevance of the experience and expertise of each role in enriching the analysis of this value chain [34].
The evaluation of technological demands was based on the report on the SIEMBRA platform (http://www.siembra.gov.co/Offers/Oferta/Reporte,accessed on 10 September 2022) and the search was made 30 June 2021.This platform is responsible for knowledge management in science, technology and innovation.It condenses information on supplies, demands, technological capabilities and future or ongoing projects [35], prioritized with the aim of focusing research needs and making efficient use of resources [36].The result of lawsuits on the platform was validated according to the perspective of the experts.The area used for querying demands is general and the results are presented descriptively.
For the data, reliability of each instrument was evaluated with the alpha (α) criterion of Cronbach [37] with the psych library [38].A cluster analysis was performed to identify whether some technologies were considered to be of greater or lesser importance by the experts, using the cluster library [39] with the K-means method [40].The result of the cluster analysis was validated with the Silhouette coefficient for internal analysis (Si > 0) [41] and the relationship between the variance between groups (BSS) and the total variance (TSS).The plotly library was used for the graphs [42].Subsequently, a Confirmatory Factor Analysis (CFA) was carried out [43] to validate the questionnaires and scales, wherein supplies were indicators and the groups of supplies were factors.The null hypothesis for this statistical analysis was that the proposed model fits the data, in which each indicator contributes to the factor additively and in equal proportion [44][45][46].
The factorial model was applied by group of technologies in a general way (all groups) to review the importance given by experts within each group and to the technology individually using the Lavaan [47], factoMineR [48] and factoextra libraries [49].For the confirmatory factor analysis, the estimator proposed by Muthén [50] "Diagonally Weighted Least Squares-DWLS" was used considering that the sample size was small and the data did not present normally.The fit of the proposed model was measured with the Comparative Fit Index (CFI > 0.95), the Tucker-Lewis index (Tucker-Lewis index-TLI > 0.90) and the mean square error (Root Mean Square Error-RMSE < 0.10) [51].Factor loadings obtained from this analysis were subsequently used to calculate the TL.The open questions were analyzed based on the frequency of use of terms, with a word cloud graph as a visualization tool, in which the size of the terms reflect the frequency with which they were used by the experts.

Evaluation of the Technological Level in Avocado Production Systems
The input generated from the expert consultation was used to design an instrument to evaluate the frequency of use of 82 technologies (in groups (i), (ii) and (iii)) by avocado producers (Appendix A).The number of technologies evaluated in each group was different: Propagation of plant material and nursery (16), Production (43) and Postharvest (23).
Producers were asked whether they knew the technology (1) or not (0) and the frequency of use, according to a Likert-type scale [33], with five response options: 5-always, 4-almost always, 3-sometimes, 2-hardly ever and 1-never.The experts stated that some of the variables related to group IV were general and therefore did not differentiate the supplies.The study, therefore, evaluated the use or non-use of the technologies.In addition, the perception of the producers regarding the technologies of this group were considered.
This research poses two scales, the value chain (expert consultation) and production systems (producers).Considering the above, it was decided to use the methodology proposed by Ramírez-Gil et al. [14], which establishes the weighting of pre-established technologies and is related to the frequency of their use at a given time, since this study evaluates the TL and not the process of technological adoption.The technologies evaluated in each group were previously assessed by experts in the production chain.The result of the factor loadings was used as a weight to calculate the TL in each group of technologies.The equation proposed by Ramírez-Gil et al. [14] was modified according to the objective of the research, evaluating the frequency of use of technologies (Equation (1)).After the TL measurement, a principal component analysis (PCA) was performed using the factoextra library [49].The PCA results were used for a cluster analysis to generate an objective classification of the producers according to the TL (low, medium and high) following the previously mentioned methodology.For this analysis, the groups of technological supplies: (i), (ii) and (iii) were included.
where "TL" is technological level, "WT" is equal to the weighting of the technology according to experts based at the factorial load and "FT" is the frequency of use of the technology.

Characteristics of Producers and Production Systems
In the instrument used for data collection, information on the characteristics of the producers and the production system was collected.The characterization variables were correlated with the TL.Correlation of the characteristics of the producers and the production systems was carried out with the Pearson coefficient and the p-value was reported by using the rstatix library [52].Subsequently, a mixed data analysis such as the combination between a PCA and multiple correspondence analysis (MCA) was used considering that the variables were heterogeneous (continuous and categorical) [53].The resulting dimensions were analyzed, adding the contributions for each variable and those with the greatest contribution (>70%) were selected to reduce their complexity.Based on the selected variables, a multinomial logistic regression was proposed to predict the TL, which has previously been used to model other social characteristics [54,55].The data matrix was divided into training data (30%) and validation data (70%) to verify overfitting.In addition, the residual deviance and AIC metrics were used to verify the fit of the model [56,57].To perform this analysis, the ipred [58] and nnet [59] libraries were used.

Consultation with Experts and Existing Demands in Avocado Production Systems
The first objective of this research included the participation of 57 experts and was the basis for validating the proposed supplies.The data collection instrument presented high reliability (α = 0.98; p < 0.05; average r = 0.35).In addition, 97% agreed with the proposal to divide the existing supplies for avocado cultivation into the four groups.
The general model showed a good fit (p < 0.05; CFI = 0.953; TLI = 0.952; RMSE = 0.052).For the entire group of technologies, the highest factorial load was in the post-harvest group and was "Obtaining by-products from the seed".On the other hand, the technology with the lowest factorial load was in the production group and was "Branch girdling".
When contrasting the calculated factor loadings of each group against the general loadings, a positive linear correlation of 0.91 was observed (Figure 1).Three clusters were found in the analysis, in which the technologies were classified according to their low, medium or high importance (Si = 0.45; BSS/TSS = 76.1%;p < 0.05) (Appendix B). "Postharvest" presented the greatest number of technologies classified as highly important.This section may be divided by subheadings.It should provide a concise and precise description of the experimental results, their interpretation and the experimental conclusions that can be drawn.
tion systems was carried out with the Pearson coefficient and the p-value was repor using the rstatix library [52].Subsequently, a mixed data analysis such as the combin between a PCA and multiple correspondence analysis (MCA) was used considerin the variables were heterogeneous (continuous and categorical) [53].The resulting d sions were analyzed, adding the contributions for each variable and those with the est contribution (>70%) were selected to reduce their complexity.Based on the se variables, a multinomial logistic regression was proposed to predict the TL, whic previously been used to model other social characteristics [54,55].The data matri divided into training data (30%) and validation data (70%) to verify overfitting.In tion, the residual deviance and AIC metrics were used to verify the fit of the model [5 To perform this analysis, the ipred [58] and nnet [59] libraries were used.

Consultation with Experts and Existing Demands in Avocado Production Systems
The first objective of this research included the participation of 57 experts an the basis for validating the proposed supplies.The data collection instrument pres high reliability (ⲁ = 0.98; p < 0.05; average r = 0.35).In addition, 97% agreed with th posal to divide the existing supplies for avocado cultivation into the four groups.
The general model showed a good fit (p < 0.05; CFI = 0.953; TLI = 0.952; RMSE = 0 For the entire group of technologies, the highest factorial load was in the post-h group and was "Obtaining by-products from the seed".On the other hand, the techn with the lowest factorial load was in the production group and was "Branch girdlin When contrasting the calculated factor loadings of each group against the g loadings, a positive linear correlation of 0.91 was observed (Figure 1).Three clusters found in the analysis, in which the technologies were classified according to thei medium or high importance (Si = 0.45; BSS/TSS = 76.1%;p < 0.05) (Appendix B). "Po vest" presented the greatest number of technologies classified as highly important section may be divided by subheadings.It should provide a concise and precise de tion of the experimental results, their interpretation and the experimental conclusion can be drawn.Factorial analyses by group showed acceptable values at the level of the statistics used.The RMSE value, however, was higher than 0.1 (Figure 2)."Propagation of plant material and nursery" had 16 technologies, where the greatest weight corresponded to "Size of the bags for the seedlings growth".On the other hand, the lowest weight was in "Plant material from a registered nursery" (p < 0.05; CFI = 0.985; TLI = 0.983; RMSE = 0.114).
taining by-products from the seed".This behavior was also observed in the general an ysis of technologies.On the other hand, the variable with the lowest factorial load in th group was "Measurement of dry matter".
The group "Management and marketing" had 22 indicators (p < 0.05; CFI = 0.993; T = 0.992; RMSE = 0.223) and presented the highest RMSE value.The indicator with t highest estimator was "Technical records" and the one with the lowest was "Digital too for information management".A total of 161 demands were found in different areas, divided by region, on the SIEM BRA platform.Demands were classified by research area, with 12 areas for avocado.T area with the highest number of demands was "Harvest, post-harvest and transformati management" followed by "Planting material and genetic improvement".The area w the highest average priority was "Environmental management and sustainability".T confirmatory factor analysis (p < 0.05; CFI = 0.994; TLI = 0.992; RMSE = 0.064) showed th the most important latent variable was "Environmental management and sustainability The "Production" factor had the highest number of technological supplies with 45 indicators (p < 0.05; CFI = 0.989; TLI = 0.989; RMSE = 0.121).In this group, the indicator with the highest estimate was "Taking leaf samples".The "Banding" technology presented the lowest value in this group, as seen in the general analysis.For this technological supply, the experts stated that this practice may present more associated risks in relation to the benefits.For the "Production" group, the experts stated that commercial houses should consider maximum residual limits but not within the framework of production.
"Postharvest" was the third group, with 23 indicators (p < 0.05; CFI = 0.979; TLI = 0.977; RMSE = 0.145), which presented a higher estimated value for the technology "Obtaining by-products from the seed".This behavior was also observed in the general analysis of technologies.On the other hand, the variable with the lowest factorial load in this group was "Measurement of dry matter".
The group "Management and marketing" had 22 indicators (p < 0.05; CFI = 0.993; TLI = 0.992; RMSE = 0.223) and presented the highest RMSE value.The indicator with the highest estimator was "Technical records" and the one with the lowest was "Digital tools for information management".
A total of 161 demands were found in different areas, divided by region, on the SIEM-BRA platform.Demands were classified by research area, with 12 areas for avocado.The area with the highest number of demands was "Harvest, post-harvest and transformation management" followed by "Planting material and genetic improvement".The area with the highest average priority was "Environmental management and sustainability".The confirmatory factor analysis (p < 0.05; CFI = 0.994; TLI = 0.992; RMSE = 0.064) showed that the most important latent variable was "Environmental management and sustainability", which coincides with the SIEMBRA platform.However, the importance assigned to other demands of the evaluated group did not coincide with the platform (Table 1).The individual opinions showed that research has not responded to demands in the sub-sector, such as planting material and genetic improvement, physiology and nutrition, soil-water management and socioeconomics.In addition, although the technologies are known or available, they have not gone through transfer processes that allow producers to use them for decision-making.This is the case with geographic information systems, zoning and environmental aspects.Likewise, there are difficulties in cultivation areas with a high slope degree, transformation aspects, generation of added value and adaptability to climate change, which are highly important demands (Figure 3).The topics with greater weight among the expert comments included "Information systems", "Zoning", "Planting material", "Georeferencing" and "Genetic improvement" (Figure 3).

Technological Level in Avocado Production Systems
The instrument for measuring the frequency of use of the technologies applied avocado producers presented high reliability (ⲁ = 0.92; p < 0.05; average r = 0.09).Wh contrasting the average importance given to each technology by the experts and the av age use of these technologies by the producers, a correlation of 0.37 was found.The "Po harvest" supplies had the lowest value of frequency of use, followed by Production a Propagation of plant material and nursery.
A correlation was found between the technological levels of each group.The produ tion and post-harvest groups were the most highly correlated (0.70), followed by the c relation between Propagation of plant material and nursery and Production (0.33).T lowest correlation was between the groups Propagation of plant material and nursery a Postharvest (0.08).Based on the PCA and cluster analyses, three technological levels w found: high (34.4%),medium (47.2%) and low (18.4%)(Si = 0.57; BSS/TSS = 80.6%, p < 0.0 The "Production" technologies presented the greatest contribution when performing t cluster analysis and those belonging to the "Propagation of plant material and nurser group had the least application by producers (Figure 4).

Technological Level in Avocado Production Systems
The instrument for measuring the frequency of use of the technologies applied to avocado producers presented high reliability (α = 0.92; p < 0.05; average r = 0.09).When contrasting the average importance given to each technology by the experts and the average use of these technologies by the producers, a correlation of 0.37 was found.The "Postharvest" supplies had the lowest value of frequency of use, followed by Production and Propagation of plant material and nursery.
A correlation was found between the technological levels of each group.The production and post-harvest groups were the most highly correlated (0.70), followed by the correlation between Propagation of plant material and nursery and Production (0.33).The lowest correlation was between the groups Propagation of plant material and nursery and Postharvest (0.08).Based on the PCA and cluster analyses, three technological levels were found: high (34.4%),medium (47.2%) and low (18.4%)(Si = 0.57; BSS/TSS = 80.6%, p < 0.05).The "Production" technologies presented the greatest contribution when performing the cluster analysis and those belonging to the "Propagation of plant material and nursery" group had the least application by producers (Figure 4).
In the TL analysis, 63% of the producers surveyed in the region of Antioquia were located in medium TL, while in Caldas 40% were located in high TL.In the other region, the number of people surveyed was lower.Notably, Cundinamarca had 82% of the producers between medium and high TL.In Risaralda, there was no low TL; all were located between medium and high TL.On the other hand, the producers surveyed in the region of Tolima were located in a high TL.For the genetic material, 55.5% of the production systems that cultivate the Hass variety had a high to medium TL, while the other varieties (Fuerte, Lorena, Semil, Booth, Reed, Colinred, Choquette, Augustus, Semil and Santana) were indistinctly distributed among the technological levels.In the TL analysis, 63% of the producers surveyed in the region of Antioquia were located in medium TL, while in Caldas 40% were located in high TL.In the other region the number of people surveyed was lower.Notably, Cundinamarca had 82% of the pro ducers between medium and high TL.In Risaralda, there was no low TL; all were located between medium and high TL.On the other hand, the producers surveyed in the region of Tolima were located in a high TL.For the genetic material, 55.5% of the production systems that cultivate the Hass variety had a high to medium TL, while the other varietie (Fuerte, Lorena, Semil, Booth, Reed, Colinred, Choquette, Augustus, Semil and Santana were indistinctly distributed among the technological levels.

Characteristics of Producers and Production Systems
It was possible to establish relationships between the technological level and socio economic characteristics of avocado producers.Significant correlations were found be tween the variables evaluated to characterize the surveyed producers (Appendix C).Fo TL, the variables academic training (0.32; p < 0.001), registration of export farm (0.28; p < 0.01), global GAP certification (0.22, p < 0.01), completion of records (0.19; p < 0.05), size o the production system (0.19; p < 0.05) and the presence of technical assistance (0.19; p < 0.05) stood out with a positive relationship.Similarly, municipalities presented a high cor relation (0.22; p < 0.05), where Manizales, Marquetalia, Pácora, Risaralda and Salamina were the ones with the highest production systems located at the medium and high tech nological levels.On the other hand, the higher the percentage of product rejection, the lower the technological level was (0.21; p < 0.05).The crop yield was higher with medium and high TL (0.21; p < 0.05).The type of employment relationship in the production system was correlated with the TL (0.21; p < 0.05), producers who employed family labor had a lower TL than those who hired external labor.

Characteristics of Producers and Production Systems
It was possible to establish relationships between the technological level and socioeconomic characteristics of avocado producers.Significant correlations were found between the variables evaluated to characterize the surveyed producers (Appendix C).For TL, the variables academic training (0.32; p < 0.001), registration of export farm (0.28; p < 0.01), global GAP certification (0.22, p < 0.01), completion of records (0.19; p < 0.05), size of the production system (0.19; p < 0.05) and the presence of technical assistance (0.19; p < 0.05) stood out with a positive relationship.Similarly, municipalities presented a high correlation (0.22; p < 0.05), where Manizales, Marquetalia, Pácora, Risaralda and Salamina were the ones with the highest production systems located at the medium and high technological levels.On the other hand, the higher the percentage of product rejection, the lower the technological level was (0.21; p < 0.05).The crop yield was higher with medium and high TL (0.21; p < 0.05).The type of employment relationship in the production system was correlated with the TL (0.21; p < 0.05), producers who employed family labor had a lower TL than those who hired external labor.
The multivariate analysis showed that 42 dimensions explained at least 70% of the variance of the collected data.These dimensions obtained the contributions for the qualitative and quantitative variables, indicating high variability and little or no self-correlation between the predictors (Figure 5).
The prioritization of variables according to the PCA and MCA methods obtained eight variables that were detrimental to the TL (Figure 5B, Equation ( 2)).The number of sources that each producer had access to for the technological aspects had an important contribution (67%).Among the categories of academic training, technological training presented the greatest contribution, as shown in the correlations (Figure 5B, Equation ( 2)).
The multivariate analysis showed that 42 dimensions explained at least 70% of the variance of the collected data.These dimensions obtained the contributions for the qualitative and quantitative variables, indicating high variability and little or no self-correlation between the predictors (Figure 5).The age of the producer presenting the greatest relevance to the TL was 18-20 years.The distance to the municipal seat of 21-30 km presented the greatest contribution for this variable type.The source of bank credit and seniority in the association of producers from 5 to 10 years presented high importance.On the other hand, the marketing channel, intermediary and the once-a-year frequency of technical assistance also had a greater contribution.From this process, equation 2 was used as a suggested model (Figure 5B, Equation ( 2)).
where "TL" is technological level, "ap" is age of the producer, "dmc" is distance to the municipality, "sc" is source of credit, "at" is academic training, "ta" is a time to be associated, "mc" is a marketing channel, "si" is a source of information on technological aspects and "fta" is the frequency of technical assistance, each multiplied by an assigned factor.The calculated model explained 53.7% of the variance of the data and an adequate adjustment (AIC = 91.8)(Table 2).When using all variables, this model explained 0.012% of the variance and presented a lower fit (AIC = 148).Among the eight variables, the source of credit had the greatest weight in the proposed model, followed by academic training.

Discussion
This study confirmed that technologies can be classified into groups of low, medium or high importance, regardless of the group (i, ii, iii, iv).The perception of the technologies used in the avocado value chain demonstrated the presence of strong and weak links.However, they varied according to the region, avocado variety and stage of the production system.
Arías García et al. [32] stated that the demands of the Hass variety are mainly concentrated in the areas of "harvest and post-harvest" and "planting and genetic improvement".This study found that the demands of the avocado sub-sector are generally linked to the production process.Supply in the harvest and post-harvest stages has been met.In contrast to the first hypothesis, despite the relevance of the management and marketing aspects, this group presented the least adjustment in the proposed model, which may be associated with characteristics of low use or little knowledge on the part of the value chain despite the importance it has gained recently [4].This indicates that the dissemination of management as a relevant topic for the sector is still in process and requires strengthening for use in agriculture [60,61].
The more important supplies were transformation of the fruit and search for added value from seeds, where a large amount of waste is generated, which may have high potential [62][63][64][65].Likewise, the size of the bag in the nursery stood out, a sensitive variable with a high influence on the quality of the plant material [66].Environmental considerations have been highly relevant given the association with the use of resources because of water and carbon footprints [67][68][69].The foregoing agrees with the consultation with experts, according to which demands related to environmental aspects presented the greatest factor loads.Like previous geographic information systems (GIS), technology for data management in space and time allows multiple applications [70][71][72][73].Additionally, genetic resources and the possibility of predicting the response of plants to biotic and abiotic factors is an advantage for avocado production systems [74,75].On the other hand, the importance of foliar analysis for decision-making and monitoring fertilization management was reflected in the expert consultation [76,77].
Although there are technologies at different stages of the chain, not all of them are considered useful or appropriate for the local context.For example, girdling is used in various fruit trees including avocado crops and this practice implies some commercial advantages [78,79].However, when testing this practice in tropical areas, the results have not been as expected [80], while dry matter is used for decision-making for harvest [81].Despite this, the value chain did not consider it relevant, though it is currently the most important harvest criterion used by producers [82].
Although some of the experts consulted were also producers, many of the evaluated technologies were not applied in production systems, leading to a gap.Singh et al. [83] found that the technological gap may be related to low prices in the market, low presence of extension agents or technology transfer.The variables with high priority according to the platform designed for this purpose in Colombia (SIEMBRA) were classified by the experts as of low importance.Therefore, the demands and their priority do not respond to the needs of the sector.
In previous experiences, it was found that considering other actors' vision allows enrichment of the research agendas.In the state of Michoacán, there was a period of closures and restrictions imposed on the exportation of avocado.After years of research and joint work between the avocado trade union and the academy, some restrictions were lifted and they are currently the main producer and exporter of Hass avocado in the world [84].Adamashvili et al. [16] affirm that the success of agriculture is at the center of technology, collaboration and knowledge, which is strengthened through public policies and financing.
The current technological state of avocado production systems allows the sub-sector to work in those technological groups with identified weaknesses.Focusing technology transfer efforts will allow the levels to be homogeneous despite varieties and geographic location.It is necessary to scale the information collected to other areas to work with regional strategies.Most farmers have some knowledge and positive attitude towards the technologies evaluated, which facilitates adoption processes [85].On the other hand, Ramírez-Gil et al. [14] found three technological levels in Hass avocado crops in Colombia, linking said level with yields, export capacities and vulnerability to the incidence of soilborne diseases.
Several authors have mentioned the importance of evaluating the characteristics of producers and production systems when considering the technological level inasmuch as these are decisive in the process of adoption or transfer [86].In this study, the TL was evaluated and a correlation and causal relationship was found with some variables that have been previously reported.Some authors have stated that age may be a factor related to the technological level since, as producers age, they tend to decrease participation in innovation spaces [87].Likewise, the distance to markets or population centers has also been related to technological characteristics, innovation, fruit quality and income [29,[88][89][90].
On the other hand, Tiruneh et al., [86] found that access to credit is a facilitating condition when deciding to adopt a technology or not, a factor that, in turn, is related to the TL.According to the present study, the determinant was the source of credit.On the other hand, the level of training is a characteristic that has been evaluated and considered positive with respect to the technological level [91].In addition, the level of training is a characteristic that decreases risk aversion in producers [92].Technology transfer may be related to the technological gap found.For instance, a report states that the gap between basic and applied research makes it difficult to increase competitiveness [85].Producer organizations are a channel for the transfer of knowledge through agreements or collective contracts, explaining the high relevance of this variable for the analysis of the TL [92].On the other hand, the marketing channels, sources of information as well as frequency of technical assistance have also been reported as influential in the TL [93].
The proposed technological groups are a basis for future research because of acceptance by experts.In addition, the technological supplies serve as a basis to evaluate the technological level of the subsector.The perception of the experts regarding the technological supplies showed that the avocado subsector needs to explore the information available to carry out transfer processes.Social and economic characteristics of the avocado production systems were found that have an effect on the technological level and that must be considered in transfer and innovation programs to achieve sustainable results.Finally, the use of the model to predict the TL is an alternative to optimize the processes of characterization and data collection in the field.

Conclusions
The methodology used in this work for analyzing and modeling the variables, identifying gaps and defining causal relationships, is new and useful for application in future research.The exploration of categorical data carried out makes it possible to reach specific decisions regarding information collection instruments.On the other hand, this work gives evidence in favor of the use of technological strategies, such as virtual forms for the diagnosis or consultation of actors of value chains in agriculture.For future replications of this work, a larger sample size is recommended in the consultation, including experts and a proportional participation of the actors of the value chain.This work is a tool for the avocado sub-sector with which research, development and innovation strategies can be guided.In addition, the technological state of avocado production systems is recognized for the first time.Supplies evaluated showed high importance; however, no correlation was found with the use of said technologies in the production systems.On the other hand, it was possible to identify the technologies that require more attention and development.The relationships found between TL and socioeconomic characteristics show the need to incorporate these variables into knowledge communication processes.Furthermore, the cohesion of actors in the avocado value chain is one of the main challenges of the sub-sector.For future work, evaluation of the adoption factors as well as the causes of the identified gap is suggested.

Figure 1 .
Figure 1.Factorial load to technologies evaluated and by group of technologies according stage of the value chain.Source: Authors.

Figure 1 .
Figure 1.Factorial load to technologies evaluated and by group of technologies according to the stage of the value chain.Source: Authors.

Figure 2 .
Figure 2. Confirmatory Factor Analysis of technological supplies consulted with experts, accordi to the stage of the value.The circular arrows indicate variances (latent variables) and covarian (center, between factors).The values in the straight arrows indicate the factor loading of each lat variable by group of technological supplies.For each group, the highest value is underlined and largest value is highlighted in red.The coding of the latent variable is found in Appendix B.

Figure 2 .
Figure 2. Confirmatory Factor Analysis of technological supplies consulted with experts, according to the stage of the value.The circular arrows indicate variances (latent variables) and covariance (center, between factors).The values in the straight arrows indicate the factor loading of each latent variable by group of technological supplies.For each group, the highest value is underlined and the largest value is highlighted in red.The coding of the latent variable is found in Appendix B.

Figure 3 .
Figure 3. Word frequency analysis of the open responses given by experts regarding the technol ical demands in avocado cultivation in Colombia.

Figure 3 .
Figure 3. Word frequency analysis of the open responses given by experts regarding the technological demands in avocado cultivation in Colombia.

Figure 4 .
Figure 4. Grouping of producers according to technological level and stage in the avocado valu chain in Colombia.

Figure 4 .
Figure 4. Grouping of producers according to technological level and stage in the avocado value chain in Colombia.

Figure 5 .
Figure 5. (A) Distribution and importance of variables of characterization of the producers according to the contribution that each of them has to explain variance of the data in the first two dimensions.Numbers were assigned to the variables to facilitate visualization. 1. Gender; 2. Age of the producer; 3. Distance to the municipal seat; 4. Land tenure; 5. Source of income; 6. Use credit; 7. Credit source; 8. Academic training; 9. Internet in production systems.10.Smartphones in production systems; 11.Computers in production systems; 12. Keep records; 13.Linked to producer association; 14.Seniority in the association of producers; 15.Production system area; 16.Crop age; 17.Years of experience in cultivation; 18. Crop yield; 19.Sale price; 20.Marketing channel; 21.Global-GAP; 22. Registration of export farm; 23.Percentage of rejection; 24.Source of information on technological aspects; 25.Causes of production reduction; 26.Recruitment of labor; 27.Payment of family labor; 28.Social security for workers; 29.Technical assistance; 30.Hiring technical assistance; 31.

Figure 5 .
Figure 5. (A) Distribution and importance of variables of characterization of the producers according to the contribution that each of them has to explain variance of the data in the first two dimensions.Numbers were assigned to the variables to facilitate visualization. 1. Gender; 2. Age of the producer; 3. Distance to the municipal seat; 4. Land tenure; 5. Source of income; 6. Use credit; 7. Credit source; 8. Academic training; 9. Internet in production systems.10.Smartphones in production systems; 11.Computers in production systems; 12. Keep records; 13.Linked to producer association; 14.Seniority in the association of producers; 15.Production system area; 16.Crop age; 17.Years of experience in cultivation; 18. Crop yield; 19.Sale price; 20.Marketing channel; 21.GlobalGAP; 22. Registration of export farm; 23.Percentage of rejection; 24.Source of information on technological aspects; 25.Causes of production reduction; 26.Recruitment of labor; 27.Payment of family labor; 28.Social security for workers; 29.Technical assistance; 30.Hiring technical assistance; 31.Frequency of Technical Assistance.(B) Socioeconomic variables with greater discrimination capacity associated with the NT of avocado producers in Colombia.
writing-original draft preparation, J.C.-Z.; writing-review and editing, J.C.-Z., J.G.R.-G.and D.B.; visualization, J.C.-Z., J.G.R.-G.and D.B.; supervision, J.G.R.-G.and D.B.; project administration, D.B.; funding acquisition, J.C.-Z.and D.B.All authors have read and agreed to the published version of the manuscript.Funding: This research was funded by Universidad Nacional de Colombia, announcement for the partial funding of Doctoral and Master's thesis projects of the Facultad de Ciencias Agrarias, Bogotá with Hermes code: 54004.Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Table 1 .
Technological demands to avocado value chain in SIEMBRA, organized by importance according to expert consultation.
* General classification according to priority to SIEMBRA.** Classification by importance according to factorial load, result of expert consultation.*** Variance of each latent variables (demand).

Table 2 .
Variables and categories used for the multinomial logistic regression model and coefficients assigned by the model.