Q) What is Statistics? How it differs from Statics?
What Type of Problems we are
going to solve by statistics? ( What , Why , Need, Problems we are going to solve ) Q) I heard that Statistics deal with drawing conclusions from data, hey but dont I have SQL for it. What I feel is when there is unstructured data I should go for statistics. (Any alternative methods to solve the problem..Was the given problem being addressed by different means or it wasnt being addressed at all) Q) How is SQL different from statistics? SQL is already there for Data Analysis? Q) What conclusions we want to draw from data and how these calculations are useful in Analytics or decision making ? Q) We didnt use Statistics in Data Warehousing and Business Intelligence Projects we have implemented so far. Do you think as a rule we should apply statistics in DW/Bi projects? Q) How we can know whether we should apply statistics in Converting Data to Information project? Q) What does Data Analysis comprise of? Q) Isnt statistics a part of mathematics, why as computer science graduates we need to learn it. Q) Are there some fixed conclusions which we need to derive from data? Q) What does Statistics offer which BI/ETL systems are not offering? ( When you think you have position yourself in different roles, student, teacher, professional, change resistant, architect, innovator etc) Q) Is Statistics a part of data analysis? Research: Systematic investigation and study of materials and sources in order to establish facts and reach new conclusions. Q) What do you mean by Experimental Research? Q) What is causality? placebo: ( A substance that has no therapeutic effect, used as a control in testing new drugs) What Type of research you are doing? what type of questions you are asking? Q) Isnt Polio vaccine a problem in medicine. How is it related to statistics? Q) What sample you should take: People who are infected with virus or people who are healthy? Q) What should be the sample size? Q) What is double- blind experiment? If no one knows who is receiving vaccine and who placebo, how results can be determined. Q) Why placebo is given at all? C: The Aim is there should not be biases.
Causality: the relationship between something that happens( polio
developed/absence) or exists and the thing that causes it (vaccine). Learnings: When I make a strong claim (this vaccine will eradicate polio), I need to carry experiments to prove it. Randomized experiments allow in helping strong claims about causality. Strong causal claims require True independent variables, random and reprentative samples andno confounds. ( which is impossible) (When we are doing an experimental research we are trying to establish whether there is causal relationships between Independent Variables and dependent Variables ). Is this statement right? Say, I developed a vaccine and find out that 10 people who were administered vaccine didnt get infected by polio. So can I now say I have developed an effective vaccine. ( No, the sample you had, it could be biased; biased in the sense that you had healthy people who had antibodies for polio ) All Medicines and drugs developed need to go through experiments/clinical trials. Sample, Population, Biases, Independent Variable, Dependent Variable, Double blind experiment Q) what is working memory? Lingo: Not real sample or representation of healthy adults : just college students Q) Is random and representative sample the same. Q) What is control group? why do we need it? What is the Treatment group? Q) What is 2 back task? As you did training, you got better at the task no that your intelligence increased or yoy are working on memory not intelligence. The People who received memory training fared better in IQ test after a month than the people who didnt receive training. So we conclude memory training increases IQ. Before training, the treatment group (people who received training) and control group (people who didnt receive training) had same IQ. Intelligence and memory capacity are fixed once you get to adulthood (hard to change IQ)
so above research was proving it to wrong, so there was buzz.
The Difference b/n control group and treatment group should be small, here control group did nothing. Why difference should be small? Concept: Strong causal claims can be made only when: a) True Independent Variables b) Random or Representative sample c) No confounds ( Control group different from treatment group) In memory training, potential confounds were people were trained on something, they felt better about themselves and they performed better in IQ test. It is true that they performed better but it was not because of memory training. Google: Potential confounds in memory training or in an experimental research. Q) What is the definition of independent variables? Experimental research depends on random selection and random assignment. Q) Is random same as arbitrary? How will you take a random sample? Q) How by selecting a random sample you can generalize or make inference about population? In gambling shop or casino, you need to check random stuff. Each subject stands for equal chance. Random assignment is done to prevent biases, we are assigning subjects randomly to conditions. (vaccine or placebo). Q) But how does random selection and random assignment help in research? Random assignment so that we have similar type of samples. Q) Do really all biases are taken care by random selection and assignment? Q) What happens when we dont take random sample? Q) What is correlation? Q) Why Psychologists are so interested in statistics? Q) What are the big 5 personality traits? How do you define personality? Correlation research is not about causality. It is more used in social sciences. Q) I have a given problem, how do we decide whether we should go for correlation research or experimental research? Personality traits are distinguishable qualities of a person and stable over time. Ocean theory of personality
Q) What is positive correlation?
Correlations is strong among the question/answers i.e. they are related. Mutual independence/interdependence of variable quantities. Different problems are solved by different type of research. Problems in social science and personality are solved more by correlation research. Big 5 Personality model is supported by correlation research. Q) Why personality research is supported by correlation research? Correlation research is more when we are trying to understand some structure. Q) define intelligence Researches which are based on studies and not experiments. Human Intelligence is a research where we also use correlation research to develop theories about human intelligence. Correlation research is used here. Intelligence is not related to academics or test taking stuff. A research to support that there is general intelligence. Experimental research talks about manipulations , variables ( experiments) and not studies or behavioural study Q) What is difference b/n study and experiment? When I come up with verbal ability and spatial ability, I should have some research to support it. The gist is: Theory of intelligence/Personality is supported by correlation research, it doesnt take in account independent and dependent variables though we talks about samples. The theory is supported by correlation where we find correlation among tests, questions and we draw conclusions. Tests helped us to know there is some spatial ability ( maths, science) and some verbal ability (history/language) but there is general ability too. Medical research: Understand the Problem. Concussions affect cognitive abilities Quasi independent variable: not random assignments to conditions, so not a true independent variable Statistics helps in what is causing it ? A correlation research is dressed up as an experimental research. we dont have true independent variables as there could be prior concussions. also we dont have true random sample, in these cases we cant claim strong causality.
Understand what type of research you are engaging in: Correlation research or Experimental research? Q) What are quasi independent variable?