Evaluation 2

From CS2610 Fall 2016
Jump to: navigation, search

Slides

Readings

Reading Critiques

Haoran Zhang 15:14:51 10/5/2016

How To Interpret Experimental Results: This article comes from a book named Doing Psychology Experiments, which is mean, this article is not only for human computer interaction research, but also for any research in the behavioral and social sciences. I believe that the human computer interaction experiments are somehow similar to the psychology experiments. Because they are all about to explore what’s in people’s mind. So this article is also helpful in the human computer interaction experiments. In this article, it talked about how to plot frequency distributions, and find out the specific distribution of experiment result, such as normal distribution, bimodal distribution, truncated distribution, and skewed distribution. In addition, it talked about some metric to explain the distributions in statistics, such as central tendency, dispersion. More than that, it also talked about how to plot relationships between variables, and how to describe the relationship. Also, how to use scatterplots to help us describing the strength of a relationship. How to interpreting results from factorial experiments by find out the main effects, and interactions. Also, we can inferential statistics, and meta-analysis. Also, we can use computer to help us interpret results. Computer are robust, and hard to make mistake if you don’t make mistake. Thus, it can help us generate graph automatically, or even statistical analysis for us. With help of computer, we can interpret experimental results more efficiently.

Tazin Afrin 14:50:47 10/7/2016

Critique of “How To Interpret Experimental Results”: In this article the author described some very basic and widely used statistical data analysis methods. The answer to our research question depends on how we want to define the terms. A practical answer is always desirable, because at the end we want to build a knowledge based system instead of just doing experiments. So to know the effect we need to analyze our data and plot the relations for better understanding. In this chapter, the author David Martin talked about how the distribution of a data set and how to plot the frequency distribution, how to plot the relationship between independent and dependent variables, how to use inferential statistics such as parametric vs non-parametric test and how to do some meta-analysis on dataset. A Gaussian or normal distribution helps us to understand if the data are skewed or not, and if that will have any impact on the result. If we want to know the most frequent grade of a class or the average grade, we need to calculate the central tendency of the dataset such as mode or mean. The mean of a dataset represents the gravity of the dataset. The standard deviation and the variance tells us how sparse is the dataset. So central tendency is the typical value of the probability distribution. These values help us to know about the dataset, but plotting these on a graph gives us insight about the dataset. For example, if we plot the histogram of the grades, we can easily interpret the results distribution of the class – like what is the most frequent grade, what is the second most frequent grade etc. A correlation co-efficient helps us to know if two series of the are correlated. For example- if the ice-cream sale in the school café is positively correlated with the math grade. Also parametric and non-parametric tests help us to take decision about statistically significance. A much higher level of analysis is meta-analysis which combines the results of many experiments. However, we must be very careful to carry out a statistical analysis. The assumption that we make while doing the experiment must be explicitly stated and we have to interpret the results based on our assumption too.

Zhenjiang Fan 17:45:31 10/9/2016

How To Interpret Experimental Results:::::::::::::::::::: The work main talks about the ways of interpreting experimental data. The first way the author brings up is plotting frequency distributions. There are different types of distributions, so choosing a proper one is very important. Even though plotting frequency distribution allows you to describe the raw experimental data in a more orderly way, but it is better to have a single number that represents how the participants in each group performed. This kind of representation is usually described as the central tendency, and there are three ways to express the central tendency: mode, median and mean. A measure of central tendency tells something useful about a distribution, but it describes only one special aspect. Dispersion is a way to describe how spread out the raw experimental data is. Dispersion is a range of values, not a single number. The reason you do an experiment is to find out if there is a relationship between the independent and dependent variables, so plotting the relationship between the two variables is a more straightforward way to evaluating the experimental data. Drawing graphs and describing functions are two ways of plotting the relationship between two variables. To further or better knowledge of the relationship between variables, we need to describe the strength of the relationship. The results of factorial experiments are more difficult to interpret, given they may have several independent variables and we need to find how each variable might impact others. Then the author goes on talking about the inferential statistics. Inferential statistics help you infer whether there is a difference between the populations.

Steven Faurie 8:34:37 10/12/2016

Steve Faurie How to Interpret Experimental Results: The chapter begins by describing several different types of distributions. It goes on to explain the concept of mean, median and mode. He refers to these as an indication of “central tendency.” To evaluate the spread of your data the author describes the concepts of range, variance and standard deviation. He goes on to describe the relationship of the standard deviation to the normal distribution and how approximately 2/3s of scores should fall within one standard deviation of the mean. The next section of the paper describes some common graphing techniques. Histograms, line graphs, and bar graphs are given as examples. Additionally, he describes the plotting of functions and describes several different classifications of functions. Scatter plots are discussed as well. The next section of the paper discusses concepts that I personally am less familiar with. It is the interpretation of multi factor experiments. The examples given describe different ways the variables interact and help give the reader some idea of how to interpret plotted data from a factorial experiment. The next section of the chapter describes statistical significance. Both 5% and 1% confidence intervals are discussed. The confidence interval is a way to measure the probability that the observations you made in your experiment were not due to chance. Most experimenters find that a 95% chance their observations are valid is good enough. I thought this chapter was a good high level overview of the ways we should conduct our experimental analysis.

Alireza Samadian Zakaria 15:55:43 10/12/2016

The 12th chapter of Doing Psychology Experiments is about interpreting experimental results. It provides some of the basics of how to analyze experimental data. The first thing that we can do is plotting data by frequency distribution diagrams. It is useful to find out whether there is a difference between conditions and we can find it out by seeing the shape of distribution. There are some shapes that are common and have names like normal, bimodal, truncated, and skewed. To describe these distributions, we can use statistics; there are two basic kinds of statics mentioned in this chapter: descriptive statics and inferential statistics. The descriptive statistic is useful for reporting some characteristics of the raw data instead of the data itself. There are some characteristics in data which can be reported like central tendency which can be a representative for the typical value of a variable. We can use one of mean, median or mode for reporting central tendency. We can also report a measure of dispersion which shows how spread put the scores are. For this purpose, we can calculate range, variance, or standard deviation, which standard deviation is more useful than the others. These measurements are good; however, in most of the times, we want to show the relationship between variables. For this purpose we can use different kinds of graphs such as bar graph, line graph, or histogram; it is also possible to describe it by some other characteristics such as correlation coefficient which is a number indicating how much our variables are dependent on each other. One of the subtle points is that these methods are for two variables: an independent variable and a dependent one; in those cases in which we have more than two variables we should consider main effects and interaction between the variables. Moreover, the author talks about inferential statistics which is a term referring to the use of statistic tests to find out whether the differences in values of the variables are made by chance. There are two kinds of tests regarding it: parametric tests and nonparametric tests. In parametric tests we have this assumption that the population distributions are normal; whereas, nonparametric tests do not assume it. At the end, the author describes meta-analysis and use of computers in general.

Anuradha Kulkarni 23:51:36 10/12/2016

How To Interpret Experimental Results: As the name suggests this chapter gives an overview of how to interpret experimental results. The chapter begins with the definition of distribution supported with examples. Then it elucidates type of distribution, graph, types of graph and the experiment variable relationship. This chapter explains the various ways of interpreting results of an experiment. The most efficient and common methods to visualize the result of experiments are frequency distribution, range and standard deviation. Bar graph are used to visualize data points that represent qualitatively different categories. Continuous variables are represented by means of Histogram or functional line. A statistical technique known as meta-analysis is used to classify the results of many experiments. Inferential statistics is used when the differences between data sets are caused by chance or by real difference in populations. The most valuable part in this chapter is about inferential statistics. This was a good paper, providing good understanding of the analyzing the experimental results.

Debarun Das 0:10:32 10/13/2016

“How To Interpret Experimental Results”::::: This paper describes in details about interpreting experimental results through data analysis. Initially, it discusses about plotting frequency distributions. It discusses about the four major types of frequency distributions – Normal, Bimodal, Truncated and Skewed. It proceeds by describing the statistics for distributions. Initially, it discusses about the different components of descriptive statistics – the mode, median, mean, standard deviation and variance. Then, it gives a brief basic idea of drawing graphs and illustrates the different types of functions – linear, curvilinear, monotonic and non-monotonic. It further discusses about scatterplot and correlation coefficient to describe the various techniques to find the strength of a relationship. Further, it discusses about interpreting results from factorial experiments. Finally, it discusses in details about the various techniques of inferential statistics, meta-analysis and the role of computers in doing statistical analysis. This chapter gives a basic idea of all the techniques that are needed for data interpretation and analysis. What I like most about this chapter is that it covers all the basic techniques that are needed to be known for a researcher who needs to interpret the data of experimental results. This knowledge of statistics and data analysis is extremely important to evaluate the significance of an experimental result.

Keren Ye 0:17:13 10/13/2016

How To Interpret Experimental Results The chapter tries to give us an understanding of the logic underlying data analysis. In general, psychologists use two basic kinds of statistics: descriptive statistics and inferential statistics. Descriptive statistic focus to describe characteristics of the data thus this method is good for explaining frequency distribution. To explain the idea, the author shows examples of analyzing central tendency and dispersion. The method is also very good for analyzing the relations between variables. Metrics such as correlation coefficients are mentioned to measure the strength of correlation. For factorial experiments, the results are often more difficult to interpret. Yet the author still provide means such as crossover interaction to solve the analyzing problem. Inferential static focus on “say something” rather than “particular samples that were”. To further explain the idea, the author gives details of several inferential tests and emphasize on the idea of what is statistically significant. Also, meta-analysis is a way. In sum, the book chapter provide us a systematic way of interpreting our experimental data. We could apply the basic concept into future study.

nannan wen 0:39:05 10/13/2016

“How To Interpret Experimental Results” by David W. Martin review: In this chapter, the author discussed some of the key high level components to analyze data in an experiment. One of the most important factor is to make the variable different with each other in a measurement of central tendency and dispersion, types and appropriateness of graphs, he also talked about some methods to interpret and describe findings. The author also explained how to experiment and evaluate human subjects on a given task or set of tasks. Provided some examples in explaining different concepts relating to experimental design and evaluation choice. He gives a list of metrics that may or may not be relevant to certain experimental designs that are standard in evaluating performance. The concepts that the author illustrates in these chapters are very helpful from my point of view, the scope of the use extend beyond the field of psychology; these methods are also applicable in many experiments in computer science, especially in human-computer interaction. When dealing with user interfaces, the peoblem tends to be on the user who uses the interfaces, so it’s extremely important and useful to do some user experiments to find out what users want and how they feels regarding to the interfaces. This techniques that described in the chapter are very helpful to me, especially the part where he explains under what conditions should we preform Z-test, T-test or an ANOVA test depending on the type of data, sample size, and number of sample sets. Another important test method is to test whether two sets of data from the same population can be called significantly different to a level of confidence. The statistical significance is measured by p value, if this value is larger than 0.05 then we think that there doesn’t exist statistical significance between these two sets of data.

Xiaozhong Zhang 2:37:50 10/13/2016

How To Interpret Experimental Results This article is also from a chapter of the book we have discussed before, Doing Psychology Experiments by David Martin. This chapter described the correct data analysis and interpretation methods. Personally speaking this chapter was important because I did have some embarrassing experiences when giving presentations but could not find the suitable expressions to describe the experiment results. The first method discussed was plotting frequency distribution. As we have covered in the last lecture, making plots was a most pretty straight forward approach to compare the distinctions of two distributions. In this chapter, several characteristic distributions and their plotting were demonstrated, such as normal distribution, bimodal distribution, truncated distribution and skewed distribution. Standard deviation and variance were discussed as statistical evaluation criteria for distribution. The concept of mode, mean and medium was also illustrated. In the end, the authors talked about inferential statistics, to prove if some data was statistically significantly different from others. That should be the most critical section from my point of view, since only the statistically verified data were meaningful, although some people were misusing this idea in daily life.