Back to search

University of Thessaly (UTH)

Statistical methods and experimental design for agreement and similarity studies between open-source software and hardware devices, and corresponding agriculture technology (agritech) market devices

Abstract

dc:description

The present PhD thesis aims to contribute to methodological approaches regarding the assessment of agreement and similarity between open-source devices and their corresponding commercials using statistical methods and experimental design. Open-source devices can be found in a vast number of fields like medicine, education, agriculture, sports, biology, research, and others. This thesis focuses on the development of a methodology applied in four open-source devices designed and developed to assess experiments in agricultural applications and research. Chapter 1 provides a brief introduction on the factors that contributed to the emergence of open-source devices. Moreover, open software and hardware are defined, and their applications in precision agriculture, smart farming and research are summarized. The open-source philosophy and the free distribution of knowledge for any individual interested in replicating, modifying, or creating such devices while sharing all the materials needed for its design is emphasized. Examples of publications and advantages of open-source devices are presented while acknowledging the need to develop a framework to assess the functionality and the reliability of their outputs. Specifically, the meanings of method comparison studies, agreement and similarity are defined and their contribution in assessing if two devices can be used interchangeably. Chapter 2 outlines the methodology used to assess the agreement and similarity between open-source and their corresponding industrial devices. Specifically, it provides an extended discussion for the usage of various models for fitting data produced by the devices under investigation and the evaluation of indices and graphs that are interpreted easily even by non-statisticians. Moreover, in case of concluding nonacceptable agreement between the device’s measurements, sources of disagreement can be identified, and even corrective actions can be applied to improve the agreement. Advantages and disadvantages of the approaches for agreement and similarity assessment are presented. This chapter also discusses the analysis followed according to the nature of the research and proper formulation of the experimental design along with key steps for the development of valid method comparison studies. Moreover, misleading, or incomplete methods to evaluate agreement and similarity between two devices are presented. Last, a summary of the methodology and the open-source hardware and software used to design and develop the devices is presented. In Chapter 3 an open-source wireless weather station that measures air temperature and humidity was designed and developed. First, its functionality was evaluated by monitoring its readings via Wi-Fi for seven days. Then, a method comparison study was designed between the device and a corresponding commercial weather station using a Thygro sensor from Symmetron. The experimental design and three methodologies that use mixed-effects models were presented along with their provision to assess the two device’s agreement. Thirty open-source devices were constructed and three commercial weather stations were used for the experiment. Three clusters were created with a ratio of ten open-source devices for one industrial. Each subject produced a trajectory of measurements over time. The experiment was terminated after seven days of continuous measurements. Since, the measurements per subject were more than 220 for some days (loss of data due to inappropriate connectivity) they were summarized in eight times zones according to the National Observatory of Athens. Two of the methodologies were applied successfully and an index was produced. The agreement is considered low, and a possible source of disagreement is the precision of the two instruments. Further research needs to be performed to include all the data occasions per day and not just the summary. Moreover, covariates such as air humidity or location of the cluster must be included in the study. Due to its modular nature, other sensors can be added and alternative wireless communication such as LoRa can be established in the open-source weather station. The advantages of using open-source weather stations are their low cost for production and modularity. In Chapter 4 a wireless LoRa open-source logger was designed and developed to aid aquaponics experiments with cultivations in perlite slabs. Specifically, the open-source logger is compatible with RS-485 soil/ soilless water content sensors. Moreover, the sensor used in the experiment can also record temperature, electrical conductivity, and pH of the slab. After testing the device and assessing its proper functionality, a method comparison study was designed to assess the agreement between the open-source device and a commercial, Handheld Meter (HH150) and SM150 Soil Moisture Sensor. The sample size was 100 perlite slabs. A measurement-error model was used to estimate the bias of the open-source device compared to the best unbiased linear predictor of the water content’s true value using the measurements taken from the commercial device. The results of the analysis reported significant differential and proportional bias but acceptable agreement for specific intervals of the measurement range, depending on the application. The advantages of using open-source devices compared to commercials is the low cost and the increased configurability of the device’s functions. Since perlite’s water content is not homogeneous along the slab, information on the water content for more than one point per slab is essential as inputs for precise water content estimation using specific algorithms. In Chapter 5 an open-source pH logger was designed and developed to assist experiments regarding pomology topics. Specifically, the device was calibrated manually using pH buffers for values of 4.01 and 7.01. Then it was tested by measuring pH from juice of citrus fruits. A waterproof temperature sensor was added to the device for temperature compensation when measuring pH. A method comparison study between the open-source device and a Hanna HI9024 Waterproof pHMeter was designed to assess their agreement. Using mixed-effects models, indices and useful graphs were produced using different methodologies. Their advantages and disadvantages for interpreting agreement were discussed. The indices reported mediocre agreement and the similarity analysis reported a fixed bias of -0.2215 pH units as source of disagreement. After using recalibration methods, the agreement improved and was considered excellent. The device is portable and can be used for in situ measurements. Moreover, the cost was substantially lower (more than 4 times lower) compared to Hanna HI9024. Future research can assess the measurements validity of the open-source device on different soil types and soilless substrates. Last, more sensors can be added to the device and more information can be retrieved from samples simultaneously while measuring for pH. In Chapter 6 an inexpensive custom-made device was designed and developed to aid experiments regarding the Mediterranean fruit fly (medfly), Ceratitis capitata. The open-source device records core, surface and external temperature of a fruit and was tested in laboratory conditions and further in field experiments of interest. For the field experiments, the device was placed on specific locations on different types of trees. Its shape and size make the device ideal to be placed in any setting the researcher wishes to investigate. The device is battery-operated and can last from 7 to 30 days depending on the logging interval. A method comparison study was designed to compare the temperature values captured by the core and the surface sensor in different conditions. Since the goal of this research was to quantify the bias between the temperature of the two locations, the differential and proportional bias was estimated using measurement-error models. The results reported a differential bias of 0.331ºC and a proportional bias of 0.982. Moreover, the precision comparison of the two sensors suggests that the core sensor maintains each temperature for more time compared to the surface sensor. Last, wireless technology such as LoRa can be added to the device with no hardware alteration. Solar panels can make the device energy-efficient and the use of a precise RTC to improve the sleep modes allows the device to function for months without replacing the battery due to its low-power consumption functionality. Chapter 7 concludes all the previous chapters along with all the findings and future work. Chapter 8 summarizes the contribution of this thesis.

Degree

thesis:*
Grantor dc:publisher
University of Thessaly (UTH)
Year dc:date
2023

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Bataka, Evmorfia
  • Μπατάκα, Ευμορφία

Subjects

dc:subject × 33

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/55393

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Bataka, Evmorfia; Μπατάκα, Ευμορφία. Statistical methods and experimental design for agreement and similarity studies between open-source software and hardware devices, and corresponding agriculture technology (agritech) market devices. University of Thessaly (UTH), 2023. http://hdl.handle.net/10442/hedi/55393